WO2022099988A1 - Object tracking method and apparatus, electronic device, and storage medium - Google Patents

Object tracking method and apparatus, electronic device, and storage medium Download PDF

Info

Publication number
WO2022099988A1
WO2022099988A1 PCT/CN2021/086020 CN2021086020W WO2022099988A1 WO 2022099988 A1 WO2022099988 A1 WO 2022099988A1 CN 2021086020 W CN2021086020 W CN 2021086020W WO 2022099988 A1 WO2022099988 A1 WO 2022099988A1
Authority
WO
WIPO (PCT)
Prior art keywords
target object
target
objects
camera
feature
Prior art date
Application number
PCT/CN2021/086020
Other languages
French (fr)
Chinese (zh)
Inventor
关英妲
周杨
刘文韬
钱晨
Original Assignee
北京市商汤科技开发有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京市商汤科技开发有限公司 filed Critical 北京市商汤科技开发有限公司
Priority to JP2022506434A priority Critical patent/JP2022552772A/en
Priority to KR1020227002278A priority patent/KR102446688B1/en
Publication of WO2022099988A1 publication Critical patent/WO2022099988A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/292—Multi-camera tracking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence

Definitions

  • the present disclosure relates to the technical field of computer vision, and in particular, to a target tracking method and device, an electronic device and a storage medium.
  • Object tracking refers to tracking the same target object (eg pedestrian, vehicle, etc.) in consecutive video frames.
  • target tracking algorithms are widely used in security and other fields, and are of great value in building a smart life.
  • ID identity document
  • the present disclosure provides a technical solution for target tracking.
  • a target tracking method comprising:
  • the distance between the target objects in the tracking area determine multiple target objects that conflict with each other;
  • a reference feature matching the current feature of the target object is determined, wherein the plurality of reference features are the corresponding reference features of the target objects before the plurality of target objects collide with each other. Describe the features extracted from multiple target objects;
  • the identification information corresponding to the reference feature matching the current feature of the target object is determined as the identification information of the target object.
  • the present disclosure by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features.
  • the identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
  • the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold
  • the extracting the current feature of the target object in response to the situation that any target object in the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects includes:
  • the conflict relationship between target objects is determined, and target tracking is performed based on this, which helps to improve the accuracy and efficiency of target tracking.
  • the method further includes:
  • any target object determine the first coordinates of the target object corresponding to multiple cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the first coordinates of the target object obtained according to the images collected by the camera. a coordinate;
  • the sub-area where the target object is located in the tracking area wherein the tracking area includes a plurality of sub-areas
  • the first coordinates of the target object corresponding to the multiple cameras are fused to obtain the fusion coordinates of the target object;
  • the method further includes:
  • the distance between the target objects in the tracking area is determined according to the fusion coordinates of the target objects in the tracking area.
  • the fusion coordinates of the target object are obtained, and the target tracking is performed based on the fusion coordinates of the target object, so that a more accurate
  • the coordinates are used for target tracking, which helps to improve the accuracy of target tracking.
  • the extracting the current feature of the target object includes:
  • Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
  • the camera for extracting the current feature of the target object is determined according to the confidence of the plurality of cameras for the sub-region where the target object is located, and the camera for extracting the current feature of the target object is determined according to the camera
  • the current feature of the target object is extracted from the collected video frame, and the current feature of the target object thus extracted can have richer visual information. Therefore, performing target matching according to the current feature of the target object extracted by this implementation method can Improve the accuracy of target matching.
  • determining the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located includes:
  • the multiple cameras determine the The camera that extracts the current features of this target object.
  • the target object is located by the plurality of cameras, and the difference between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the plurality of cameras.
  • the overlapping information of the target object is determined, the camera used for extracting the current feature of the target object is determined, and the current feature of the target object is extracted according to the video frame collected by the camera used for extracting the current feature of the target object, and the target object extracted from this
  • the current feature of the target object can have richer visual information, therefore, target matching is performed according to the current feature of the target object extracted in this implementation manner, which can further improve the accuracy of target matching.
  • the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
  • the determination is based on the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras.
  • the camera used to extract the current features of the target object including:
  • the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition
  • the intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
  • the camera with the highest confidence in the sub-region where the target object is located is determined as the camera used for extracting the target object among the cameras whose currently collected video frames satisfy the overlapping condition.
  • the current feature of the camera, and the current feature of the target object is extracted according to the video frame collected by the camera used to extract the current feature of the target object.
  • the current feature of the target object thus extracted can have richer visual information. Therefore, , and perform target matching according to the current feature of the target object extracted by this implementation, which can further improve the accuracy of target matching.
  • the method further includes:
  • the confidence level of the camera for the plurality of sub-areas is determined.
  • the confidence of the camera for the plurality of sub-regions is determined according to the distance between the camera and the plurality of sub-regions of the tracking area, and based on the This determined confidence level performs coordinate fusion and/or determines the camera used to extract the current features of the target object, thereby helping to improve the accuracy of target tracking.
  • the method further includes:
  • the confidence level of the camera with respect to the plurality of sub-regions is adjusted.
  • any camera in the plurality of cameras according to the average distance between the target objects in the video frames collected by the camera, dynamically adjust the confidence level of the camera with respect to the plurality of sub-regions, and Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
  • the determining a reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects includes:
  • a reference feature with the greatest similarity to the current feature of the target object is determined from a plurality of reference features corresponding to the plurality of target objects, and the reference feature with the highest similarity is responsive to the reference feature with the highest similarity with the target object.
  • the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object, so that the target object and multiple
  • a reference feature matching the current feature of the target object is searched, thereby helping to improve the search efficiency of the target object.
  • the speed and accuracy of the current feature matching the benchmark feature is performed.
  • the method further includes:
  • the reference feature corresponding to the conflicting target object is determined, and the reference feature matching the current feature of the target object is determined, so that the determined reference feature can be further improved on the premise of improving the speed of finding the reference feature matching the current feature of the target object.
  • the accuracy of the fiducial feature matching the current features of this target object is determined, and the accuracy of the fiducial feature matching the current features of this target object.
  • the method further includes:
  • the last moment does not conflict with other target objects and does not conflict with all target objects.
  • the target object matched by other target objects at the current moment is determined to be the remaining target objects at the previous moment; situation, the target object in the conflict zone to which the target object belongs at the last moment, and the target object that does not match with other target objects at the current moment, is determined as the remaining target object at the last moment;
  • the target object with the closest distance to the target object at the current moment is determined as the target object matching the target object at the current moment, and the matching target
  • the identification information of the object is used as the identification information of the target object.
  • the amount of calculation and the time overhead can be reduced, and the real-time requirement of target tracking can be met.
  • a target tracking device comprising:
  • a first determining part configured to determine a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area
  • an extraction part configured to extract the current feature of the target object in response to a situation in which any one of the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects;
  • the second determination part is configured to determine, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature that matches the current feature of the target object, wherein the plurality of reference features are in the plurality of targets Before the objects collide with each other, the features extracted from the multiple target objects respectively;
  • the third determining part is configured to determine, as the identification information of the target object, the identification information corresponding to the reference feature matched with the current feature of the target object.
  • an electronic device comprising: one or more processors; a memory for storing executable instructions; wherein the one or more processors are configured to invoke the memory storage executable instructions to perform the above method.
  • a computer-readable storage medium having computer program instructions stored thereon, the computer program instructions implementing the above method when executed by a processor.
  • a computer program including computer readable code, which when executed in an electronic device, implements the above method when executed by a processor in the electronic device.
  • the present disclosure by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features.
  • the identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
  • FIG. 1 shows a flowchart of a target tracking method provided by an embodiment of the present disclosure.
  • FIG. 2 shows a schematic diagram of a plurality of target objects colliding with each other at time t.
  • FIG. 3 shows a schematic diagram of a plurality of target objects colliding with each other at time t+1.
  • FIG. 4 shows a schematic diagram of multiple target objects and multiple reference features corresponding to the multiple target objects in an embodiment of the present disclosure.
  • FIG. 5 shows a schematic diagram of placing markers in a tracking area in an embodiment of the present disclosure.
  • FIG. 6 shows a block diagram of a target tracking apparatus provided by an embodiment of the present disclosure.
  • FIG. 7 shows a block diagram of an electronic device 800 provided by an embodiment of the present disclosure.
  • FIG. 8 shows a block diagram of an electronic device 1900 provided by an embodiment of the present disclosure.
  • FIG. 1 shows a flowchart of a target tracking method provided by an embodiment of the present disclosure.
  • the execution subject of the target tracking method may be a target tracking device.
  • the target tracking method may be performed by a terminal device or a server or other processing device.
  • the terminal device may be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, or a wearable devices, etc.
  • the target tracking method may be implemented by a processor invoking computer-readable instructions stored in a memory. As shown in FIG. 1 , the target tracking method includes steps S11 to S14.
  • step S11 according to the distance between the target objects in the tracking area, a plurality of target objects that conflict with each other are determined.
  • the embodiments of the present disclosure can be applied to application scenarios such as intelligent monitoring, intelligent scene analysis, security protection, target detection, and target tracking.
  • the tracking area represents an area where target tracking needs to be performed.
  • the tracking area in the embodiment of the present disclosure may be a relatively closed area or a relatively open area.
  • the tracking area may be a stadium, a mall, a classroom, and the like.
  • the target object represents the object that needs to be tracked.
  • the target object can be any object that needs to be tracked, such as a pedestrian, a vehicle, an athlete in a sports field (eg, a player in a football field).
  • target objects that conflict with each other may represent target objects that are closer to each other.
  • the distance between the target objects in the tracking area can be determined respectively, so as to determine the conflict between the target objects at the moment corresponding to the video frame.
  • the video frames collected by the camera may not be analyzed frame by frame.
  • the distance between the target objects in the tracking area may be determined every several video frames, so as to determine the distance between the target objects at the moment corresponding to the video frame. conflict situation.
  • FIG. 2 shows a schematic diagram of a plurality of target objects colliding with each other at time t. As shown in FIG. 2 , at time t, the target objects O 1 , O 4 and O 3 conflict with each other, the target objects O 2 and O 5 conflict with each other, and the target objects O 6 and O 7 conflict with each other.
  • multiple target objects that conflict with each other may be added to the same conflict area, that is, multiple target objects that conflict with each other may be considered to be in the same conflict area.
  • the conflict area may be a virtual area, and the target objects in any conflict area conflict with each other.
  • the conflict zone 1 includes target objects O 1 , O 4 and O 3
  • the conflict zone 2 includes target objects O 2 and O 5
  • the conflict zone 3 includes target objects O 6 and O 7 .
  • step S12 in response to the situation that any one of the multiple target objects is no longer in conflict with other target objects of the multiple target objects, the current feature of the target object is extracted.
  • FIG. 3 shows a schematic diagram of a plurality of target objects colliding with each other at time t+1.
  • conflict area 1 includes target objects O 1 and O 4
  • conflict area 2 includes target objects O 2 and O 5
  • conflict area 3 includes target objects O 6 and O 7 . That is, at time t+1, the target objects O 1 and O 4 conflict with each other, the target objects O 2 and O 5 conflict with each other, and the target objects O 6 and O 7 conflict with each other.
  • the target object O 3 is no longer in conflict with O 1 and O 4 ; that is, all the target objects in the scene converge dynamically at any time. Classes into multiple conflict zones.
  • the time interval between two adjacent moments may be equal to the inverse of the frame rate at which the video frame is captured by the camera.
  • the time interval between two adjacent moments may be greater than the inverse of the frame rate at which the camera captures video frames, for example, it may be equal to H times the inverse of the frame rate at which the camera captures video frames, wherein , H is an integer greater than 1.
  • the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold; the response In a situation where any one of the multiple target objects is no longer in conflict with other target objects in the multiple target objects, extracting the current feature of the target object includes: for any one of the multiple target objects The target object, in response to the situation that the distances between the target object and other target objects in the plurality of target objects are all greater than the distance threshold, extract the current feature of the target object.
  • a noise-based density clustering method (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) can be used to cluster target objects whose distances from other target objects are less than a distance threshold to form conflict zone.
  • DBSCAN Density-Based Spatial Clustering of Applications with Noise
  • the distance between any two target objects can be calculated, and if the distance between the two target objects is less than or equal to a distance threshold, it can be determined that the two target objects conflict with each other. If the distance between another target object and at least one of the two target objects is less than or equal to the distance threshold, it may be determined that the three target objects conflict with each other. For example, at time t, if the distance between the target object O1 and the target object O4 is less than or equal to the distance threshold, it can be determined that the target object O1 and the target object O4 conflict. If the distance between the target object O 3 and the target object O 4 is less than or equal to the distance threshold, it can be determined that the target objects O 1 , O 4 and O 3 conflict with each other.
  • the target objects when judging whether the target objects conflict with each other, it may only consider whether the distance between the positions of the target objects (for example, fusion coordinates) is less than or equal to the distance threshold, for example, if any two target objects If the distance between the positions is less than or equal to the distance threshold, it can be determined that the two target objects collide with each other.
  • the distance between the positions of the target objects for example, fusion coordinates
  • the distance between the positions of the target objects may be considered, for example, if any two targets If the distance between the positions of the objects is less than or equal to the first distance threshold, and the distance between the visual features of the two target objects is less than or equal to the second distance threshold, it can be determined that the two target objects conflict with each other.
  • the distance between the visual features of the target object may represent the similarity between the visual features of the target object.
  • the conflict relationship between target objects is determined, and target tracking is performed based on this, which helps to improve the accuracy and efficiency of target tracking.
  • the distance between any target object among the multiple target objects that conflict with each other and all other target objects among the multiple target objects is less than or equal to a distance threshold; the In response to a situation in which any one of the plurality of target objects no longer conflicts with other target objects in the plurality of target objects, extracting the current feature of the target object includes: for any one of the plurality of target objects; A target object, when the distance between the target object and any target object among the plurality of target objects is greater than the distance threshold, extract the current feature of the target object.
  • the extracting the current feature of the target object includes: determining a sub-area where the target object is located in the tracking area; Determine the camera used for extracting the current feature of the target object; according to the video frame collected by the camera used for extracting the current feature of the target object, extract the current feature of the target object.
  • the tracking area may include at least one sub-area, for example, the tracking area may be divided into a plurality of sub-areas.
  • the tracking area may be divided into a plurality of sub-areas.
  • the camera for extracting the current feature of the target object is determined according to the confidence of the plurality of cameras for the sub-region where the target object is located, and the camera for extracting the current feature of the target object is determined according to the camera
  • the current feature of the target object is extracted from the collected video frame, and the current feature of the target object thus extracted can have richer visual information. Therefore, performing target matching according to the current feature of the target object extracted by this implementation method can Improve the accuracy of target matching.
  • the determining a camera for extracting the current feature of the target object according to the confidence of the sub-region where the target object is located by the plurality of cameras includes: according to the plurality of cameras for the target object The confidence level of the sub-region where it is located, and the overlapping information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the method used to extract the current feature of the target object. Camera.
  • the confidence of the sub-region where the target object is located by the plurality of cameras, and the difference between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the plurality of cameras Overlapping information, determine the camera used for extracting the current feature of the target object, and extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object, and extract the current feature of the target object.
  • the current feature can have richer visual information, therefore, performing target matching according to the current feature of the target object extracted from this example can further improve the accuracy of target matching.
  • the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects; the confidence level of the sub-region where the target object is located according to the plurality of cameras, and the The overlapping information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by multiple cameras, and determining the camera used to extract the current feature of the target object, including: from the cameras that meet the overlapping conditions, The camera with the highest confidence for the sub-region where the target object is located is determined as the camera used for extracting the current feature of the target object, wherein the overlapping condition indicates that the target object is in the currently collected video frame.
  • the intersection ratios of the detection frame and detection frames of other target objects are all smaller than a predetermined threshold.
  • the sub-area where the target object is located is the sub-area A 1
  • the multiple cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4
  • the confidence level of the camera C 1 for the sub-area A 1 is Z 11.
  • the confidence level of the camera C2 for the sub-region A1 is Z 21
  • the confidence level of the camera C3 for the sub - region A 1 is Z 31
  • the confidence level of the camera C 4 for the sub -region A 1 is Z 41
  • the target object If there is no other detection frame whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold in the video frame currently collected by the camera C1 , that is, in the video frame currently collected by the camera C1 , the target object If the intersection ratio between the detection frame of the object and the detection frames of other target objects is smaller than the predetermined threshold, the current feature of the target object is extracted according to the video frame collected by the camera C1 . If in the video frame currently collected by the camera C1 , there are other detection frames whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold, then it is judged whether there is a video frame currently collected by the camera C2 that is related to the target object.
  • intersection ratio of the detection frame of the target object is greater than or equal to other detection frames with a predetermined threshold, that is, to determine whether the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera C 2 is not. are less than the predetermined threshold.
  • the target object If there is no other detection frame whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold in the video frame currently collected by the camera C2, that is, in the video frame currently collected by the camera C2 , the target object The intersection ratio of the detection frame of the target object and the detection frame of other target objects is smaller than the predetermined threshold, then according to the video frame collected by the camera C 2 , the current feature of the target object is extracted, and so on; In the video frame currently collected by the camera with high confidence, if the target object is occluded, you can switch to the camera with the next highest confidence, and so on; Under the condition that the camera is not blocked, it is a camera with high confidence as much as possible, which reduces the missed detection rate and improves the accuracy of target matching.
  • the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera can indicate that the target object is affected by other target objects.
  • the proportion of target object occlusion The larger the intersection ratio between the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera, the greater the proportion of the target object being occluded by other target objects; the video currently collected by the camera is larger. The smaller the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the frame, the smaller the proportion of the target object being occluded by other target objects.
  • the camera with the highest confidence for the sub-region where the target object is located among the cameras whose currently collected video frames meet the overlapping condition, is determined as the camera for extracting the current feature of the target object , and extract the current feature of the target object according to the video frame collected by the camera used to extract the current feature of the target object, so that the current feature of the target object extracted can have richer visual information. Therefore, according to this example The extracted current features of the target object are used for target matching, which can further improve the accuracy of target matching.
  • step S13 based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is determined, wherein the plurality of reference features are when the plurality of target objects conflict with each other Before, the features extracted for the multiple target objects are respectively.
  • FIG. 4 shows a schematic diagram of multiple target objects and multiple reference features corresponding to the multiple target objects in an embodiment of the present disclosure.
  • the plurality of target objects include target object O 1 , target object O 2 , target object O 3 and target object O 4 , wherein the reference feature of target object O 1 is reference feature F 01 , and the target object The reference feature of O 2 is the reference feature F 02 , the reference feature of the target object O 3 is the reference feature F 03 , and the reference feature of the target object O 4 is the reference feature F 04 .
  • the reference feature of the target object may be extracted in the first frame, and the reference feature of the target object may be stored.
  • the fiducial features of all target objects can be extracted in the first frame, and the fiducial features of all target objects can be stored.
  • the reference features of these target objects can be extracted in subsequent video frames to obtain higher-quality reference features containing richer visual information of the target objects.
  • the correspondence between the identification information of the target object and the reference feature of the target object may be stored.
  • the manner of determining the camera for extracting the reference feature of the target object is similar to the manner in which the camera for extracting the current feature of the target object is determined above, and details are not described herein again.
  • the current feature of the target object can be updated according to the extracted current feature of the target object.
  • the fiducial features of the target object to improve the accuracy of target matching based on the stored fiducial features.
  • the stored reference feature of the target object may be weighted with the current feature of the target object to obtain a new reference feature.
  • the reference feature of the target object may not be updated to reduce the amount of computation.
  • the feature of the target object can be extracted by a ReID (person Re-IDentificaion, pedestrian re-identification) module.
  • a ReID person Re-IDentificaion, pedestrian re-identification
  • the reference feature and/or the current feature of the target object can be extracted by the ReID module.
  • the ReID module can be implemented using a convolutional neural network.
  • other feature extraction methods can also be used to extract the features of the target object, as long as the extracted features of the target object can reflect the visual information of the target object.
  • the determining, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature that matches the current feature of the target object includes: selecting from a plurality of reference features corresponding to the plurality of target objects Among the reference features, the reference feature with the greatest similarity with the current feature of the target object is determined; in response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold , and the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object.
  • the similarity may be cosine similarity or the like.
  • the conflict area 1 includes target objects O 1 , O 4 and O 3
  • the reference features corresponding to the conflict area 1 include the reference feature F 01 of the target object O 1 and the reference feature F 04 of the target object O 4 and the reference feature F 03 of the target object O 3 .
  • the reference feature F 03 has the greatest similarity with the current feature F 1n of the target object, and the similarity between the reference feature F 03 and the current feature F 1n of the target object is greater than or equal to similarity If the degree threshold is set, the reference feature F 03 can be determined as the reference feature matching the current feature of the target object.
  • a reference feature with the greatest similarity to the current feature of the target object is determined from a plurality of reference features corresponding to the plurality of target objects, and the reference feature with the highest similarity is responsive to the reference feature with the highest similarity with the target object.
  • the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object, so that the target object and multiple
  • a reference feature matching the current feature of the target object is searched, thereby helping to improve the search efficiency of the target object.
  • the speed and accuracy of the current feature matching the benchmark feature is performed.
  • the method further includes: in response to the reference feature with the greatest similarity and the current feature of the target object When the similarity between them is less than the similarity threshold, based on the reference feature corresponding to the target object that does not conflict with the target object and conflicts with any other target object, the reference feature matching the current feature of the target object is determined.
  • the reference features corresponding to other conflict areas (that is, not The reference feature corresponding to the target object that conflicts with the target object but conflicts with other target objects) is searched for the reference feature that matches the current feature of the target object. For example, if the similarity between the multiple reference features corresponding to the multiple target objects and the current feature of the target object is less than the similarity threshold, it may be determined that among the multiple reference features corresponding to the multiple target objects, find the There are no fiducial features that match the current features of this target object.
  • the reference feature corresponding to the target object is determined, and the reference feature matching the current feature of the target object is determined, so as to improve the speed of finding the reference feature matching the current feature of the target object.
  • a reference feature matching the current feature of the target object cannot be found in all conflict regions, a reference feature matching the current feature of the target object can be searched in the non-conflict region.
  • the determining, based on the plurality of reference features corresponding to the plurality of target objects, the reference features matching the current features of the target objects includes: determining the reference features corresponding to the plurality of target objects. Among the plurality of reference features, the reference feature with the greatest similarity with the current feature of the target object; the reference feature with the greatest similarity is determined as the reference feature matched with the current feature of the target object.
  • step S14 the identification information corresponding to the reference feature matching the current feature of the target object is determined as the identification information of the target object.
  • the identification information of the target object may be information that can be used to uniquely identify the target object, for example, may be an ID, a serial number, a name, and the like. For example, if the tracking area is a football field and the target object includes players on the football field, the identification information of the target object may be the team to which the player belongs and the jersey number.
  • O 3 may be determined as the identification information of the target object.
  • the target tracking result corresponding to any target object may also include position information of the target object.
  • the behavior of the target object in the tracking area can be analyzed. For example, if the tracking area is a football field, and the target object includes players on the football field, then according to the identification information of the players (such as the team and jersey numbers) and the position information of the players in at least one video frame, the behavior of the players can be determined. Analysis, such as analyzing whether the player is offside, etc.
  • the present disclosure by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features.
  • the identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
  • the embodiments of the present disclosure can be applied to complex target tracking scenarios, for example, the target objects in the tracking area have different sizes, there are small-sized target objects in the tracking area that are difficult to detect, the target objects in the tracking area are dense, and the target objects are too small to be detected. Target tracking scenarios such as similar appearance, high motion complexity of the target object, and severe occlusion.
  • the method further includes: acquiring the first coordinates of the marker; for any camera in the plurality of cameras, obtaining the marker's coordinates according to the camera The second coordinate, and the first coordinate of the marker, determine the transformation matrix corresponding to the camera.
  • markers when calibrating the plurality of cameras, markers may be placed in the tracking area at a certain density first.
  • FIG. 5 shows a schematic diagram of placing markers in a tracking area in an embodiment of the present disclosure.
  • the tracking area is a football field and the marker is a white metal sheet.
  • markers placed according to a certain density can be of different heights, different sizes, and different colors, which are not limited here.
  • the first coordinates may be coordinates in a world coordinate system
  • the second coordinates may be coordinates in a pixel coordinate system
  • the transformation matrix may be a homography matrix. Since the target object moves in the same plane, formula (1) can be used to determine the transformation matrix H 3 ⁇ 3 :
  • (X w , Y ⁇ ) represents the first coordinate of the marker, for example, it can be the coordinate of the marker in the world coordinate system;
  • (u, v) represents the second coordinate of the marker obtained by the camera, for example, it can be the marker The coordinates of the object in the pixel coordinate system.
  • the degree of freedom of the homography matrix H 3 ⁇ 3 is 8, and theoretically, the homography matrix can be solved by using the feature points corresponding to 4 markers.
  • the least squares method can be used, outliers can be filtered, and the homography matrix can be obtained by using the checkerboard calibration method.
  • the homography matrix thus obtained can remove the distortion in the image captured by the camera, and can improve the accuracy of the determined first coordinate.
  • the camera external parameters can be obtained through Direct Linear Transformation (DLT) and Singular Value Decomposition (SVD), and the camera internal parameters can be obtained through checkerboard calibration to remove image distortion.
  • DLT Direct Linear Transformation
  • Singular Value Decomposition Singular Value Decomposition
  • the second coordinates of the target object obtained by the multiple cameras can be transformed into a unified coordinate system, for example, by converting In the world coordinate system, the first coordinates of the target object corresponding to each of the plurality of cameras are obtained, thereby facilitating subsequent tracking of the target object.
  • the method further includes: for any target object, determining that the target object corresponds to the first coordinates of the plurality of cameras; according to the first coordinates of the target object corresponding to the plurality of cameras , determine the sub-area where the target object is located in the tracking area, wherein the tracking area includes multiple sub-areas; The object is fused corresponding to the first coordinates of the plurality of cameras, and the fusion coordinates of the target object are obtained; before the plurality of conflicting target objects are determined according to the distance between the target objects in the tracking area, the method further includes: The distance between the target objects in the tracking area is determined according to the fusion coordinates of the target objects in the tracking area.
  • any target object corresponds to the first coordinates of any camera, and may represent the first coordinates of the target object obtained according to the images collected by the camera.
  • the first coordinates may be coordinates in a first coordinate system, for example, the first coordinate system may be a world coordinate system or other virtual coordinate systems.
  • the fusion coordinates and the first coordinates may be coordinates in the same coordinate system, for example, the fusion coordinates and the first coordinates may both be coordinates in the world coordinate system.
  • the coordinate range of each sub-area in the tracking area in the first coordinate system is predetermined.
  • the coordinate range of any sub-region in the first coordinate system can be represented by the coordinates of the four vertices of the sub-region in the first coordinate system, or, the sub-region in the first coordinate system
  • the coordinate range below can be represented by the coordinates of the top-left corner vertex of the sub-region in the first coordinate system and the width and height of the sub-region.
  • the coordinate range of any sub-region under the first coordinate system may also be represented by other manners, which is not limited here.
  • the shape of any sub-region may not be a rectangle, for example, a triangle or the like. Different sub-areas in the tracking area can be the same size or different.
  • the sub-region where the target object is located in the tracking region can be determined .
  • determining the sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras includes: according to the target object corresponding to the The first coordinates of any camera in the plurality of cameras, and the coordinate range of the sub-region in the first coordinate system, determine the candidate sub-region where the target object is located in the tracking region; the candidate sub-region with the highest number of votes is determined.
  • the area is determined as the sub-area where the target object is located.
  • the plurality of cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4 ; according to the first coordinate of the target object corresponding to the camera C 1 , determine where the target object is located in the tracking area
  • the candidate sub-region is sub-region A 1 ; according to the first coordinate of the target object corresponding to the camera C 2 , it is determined that the candidate sub-region where the target object is located in the tracking region is sub-region A 1 ; according to the target object Corresponding to the first coordinate of the camera C3, it is determined that the candidate sub-region where the target object is located in the tracking area is the sub-region A2 ; according to the first coordinate of the target object corresponding to the camera C4 , determine the target object
  • the candidate sub-area located in the tracking area is the sub-area A 1 ; then, the number of votes for the candidate sub-area A 1 is 3, and the number of votes for the candidate sub-area A 2 is 1. Therefore, the candidate sub-area A can be 1 is determined
  • determining the sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras includes: according to the target object corresponding to the The first coordinates of the multiple cameras are determined, and the average value of the first coordinates of the target object corresponding to the multiple cameras is determined; according to the coordinate range of the sub-region in the first coordinate system, the average value of the first coordinates is determined The sub-area where the target object is located, and the sub-area where the average value of the first coordinates is located is determined as the sub-area where the target object is located in the tracking area.
  • the first coordinates of the target object corresponding to all or part of the cameras in the plurality of cameras may be fused according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located, Get the fusion coordinates of the target object.
  • the multiple cameras include camera C 1 , camera C 2 , camera C 3 and camera C 4 , the target object is located in sub-region A 1 , the confidence level of camera C 1 for sub-region A 1 is Z 11 , and camera C 2
  • the confidence level for sub-region A 1 is Z 21
  • the confidence level for camera C 3 for sub-region A 1 is Z 31
  • the confidence level for camera C 4 for sub-region A 1 is Z 41
  • the target object corresponds to camera C
  • the first coordinate of 1 is P w1
  • the first coordinate of the target object corresponding to the camera C 2 is P w2
  • the first coordinate of the target object corresponding to the camera C 3 is P w3
  • the first coordinate is P w4
  • the fusion coordinate of the target object can be
  • a part of the cameras with higher confidence in the sub-region where the target object is located may be selected to determine the fusion coordinates of the target object.
  • Z 11 the confidence level of camera C 1 for
  • the fusion coordinates of the target object are obtained, and the target tracking is performed based on the fusion coordinates of the target object, so that a more accurate
  • the coordinates are used for target tracking, which helps to improve the accuracy of target tracking.
  • the target is determined according to the second coordinates of the target object obtained by the camera and the transformation matrix corresponding to the camera The object corresponds to the camera's first coordinates.
  • the second coordinate of the target object obtained by the camera may be the coordinates of the target object in a pixel coordinate system corresponding to the camera. That is, the second coordinates may represent coordinates in the pixel coordinate system.
  • the target object corresponds to the first coordinates of the camera, and represents the coordinates of the target object in the world coordinate system determined according to the second coordinates of the target object obtained by the camera.
  • the transformation matrix corresponding to the camera may be a homography matrix.
  • the second coordinate of the target object obtained by the camera may also be the coordinates of the target object in a coordinate system such as an image coordinate system corresponding to the camera. That is, the second coordinates may also be coordinates in a coordinate system such as an image coordinate system.
  • target detection may be performed on an image collected by the camera, a detection frame of the target object in the image may be determined, and according to the position of the detection frame, the second coordinates of the target object obtained by the camera may be determined .
  • any point on the detection frame or any point inside the detection frame may be used as the second coordinate of the target object.
  • the midpoint of the bottom edge of the detection frame can be used as the second coordinate of the target object.
  • the transformation matrix corresponding to any camera in the plurality of cameras is H
  • the second coordinate of any target object obtained by the camera is P u
  • the first coordinate of the target object corresponding to the camera can use the formula (2) Determine:
  • the method further includes: for any camera in the plurality of cameras, according to the distance between the camera and the plurality of sub-regions of the tracking area, determine that the camera is aimed at the tracking area. Confidence for multiple subregions.
  • the confidence levels of the multiple cameras with respect to the multiple sub-areas may be determined according to the distances between the multiple cameras and the multiple sub-areas of the tracking area.
  • the confidence of any camera with respect to any sub-region is negatively correlated with the distance between the camera and the sub-region. That is, the larger the distance between the camera and the sub-area, the lower the confidence of the camera in the sub-area; the smaller the distance between the camera and the sub-area, the lower the confidence of the camera in the sub-area higher degree.
  • the confidence Z 11 of the camera C 1 for the sub-area A 1 may be greater than that of the camera C 2 Confidence Z 21 for sub-region A 1 .
  • the confidence Z 11 of the camera C 1 for the sub-area A 1 may be greater than that of the camera C 1 Confidence Z 12 for sub-region A 2 .
  • the confidence level of the camera for the plurality of sub-areas is determined according to the distance between the camera and the plurality of sub-areas of the tracking area, The distance between the camera and the sub-area is thus taken into consideration for the confidence of the camera for the sub-area. Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
  • the method further includes: for any camera in the plurality of cameras, according to the camera The average distance between the target objects in the collected video frames is used to adjust the confidence of the camera for the multiple sub-regions.
  • the confidence of the camera for the multiple sub-regions may be continuously adjusted according to the average distance between the target objects in the video frames collected by the camera.
  • the confidence of the camera with respect to the plurality of sub-regions may be adjusted at a preset frequency.
  • the preset frequency may be 1 second.
  • a weighting process may be performed on the confidences of multiple cameras with respect to the sub-region where the target object is located, and then according to the weighted confidences of the multiple cameras with respect to the sub-region where the target object is located, The first coordinates of the target object corresponding to the multiple cameras are fused.
  • the confidence level of the camera with respect to the plurality of sub-regions is adjusted.
  • the multiple cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4 , the average distance between target objects in the video frames collected by the camera C 1 is D 1 , and the video frames collected by the camera C 2
  • the average distance between the target objects is D 2
  • the average distance between the target objects in the video frame collected by the camera C 3 is D 3
  • the average distance between the target objects in the video frame collected by the camera C 4 is D 4
  • D 1 >D 2 >D 3 >D 4 the confidence of the camera C 1 for each of the sub-regions in the plurality of sub-regions can be increased by B 1
  • the camera C 2 for each of the sub-regions in the plurality of sub-regions.
  • B 2 The confidence of the region is increased by B 2
  • B 3 the confidence of the camera C 3 for each of the sub-regions is increased by B 3
  • B 3 the confidence of the camera C 4 for each of the sub-regions is kept unchanged.
  • the smaller the average distance between the target objects in the video frame collected by the camera the smaller the average distance between the target objects in the video frame collected by the camera.
  • the larger the average distance between the target objects in the video frames collected by the camera the less likely the target objects in the video frames collected by the camera are occluded from each other, which is conducive to the feature extraction of the target objects. This facilitates target tracking.
  • any camera in the plurality of cameras according to the average distance between the target objects in the video frames collected by the camera, dynamically adjust the confidence level of the camera with respect to the plurality of sub-regions, and Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
  • the method further includes: for any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment
  • the target object that does not conflict with other target objects at the last moment and does not match other target objects at the current moment is determined as the remaining target objects at the last moment; or, in response to the target object being in the current moment
  • the target object that does not match with other target objects at the current time in the conflict zone to which the target object belongs to the previous time is determined as The remaining target objects at the last moment; from the remaining target objects at the last moment, the target object with the closest distance to the target object at the current moment is determined to match the target object at the current moment
  • the target object, and the identification information of the matching target object is used as the identification information of the target object.
  • a certain target object does not conflict with other target objects at the current moment and the last moment of the current moment, and there are 15 target objects that do not conflict with other target objects at the last moment, which are target objects 0 1 to O 15 .
  • the target objects O 1 and O 2 in the previous moment have been matched with other target objects at the current moment, that is, the identification information O 1 and O 2 have been used as the identification information of the target objects at the current moment
  • the target object O 3 To O 15 is determined as the remaining target object at the previous moment.
  • the target object with the closest distance to the target object at the current moment is determined as the target object at the current moment.
  • the target object that matches this target object For example, if among the remaining target objects O 3 to O 15 at the previous moment, O 6 is the closest to the target object at the current moment, then the target object O 6 at the previous moment is determined to match the target object at the current moment. the target object, and the identification information of the target object at the current moment is determined as O 6 .
  • a certain target object conflicts with the same other target objects at the current moment and the last moment.
  • the conflict area to which the target object belongs at the last moment includes target object O 1 , target object O 2 , The target object O 3 and the target object O 4 . If in the conflict zone to which the target object belongs at the last moment, the target object O1 and the target object O2 have been matched with other target objects at the current moment, that is, the identification information O1 and O2 have been used as the current moment's The identification information of the target object, then the target object O 3 and the target object O 4 can be determined as the remaining target objects at the previous moment.
  • the target object with the closest distance to the target object at the current moment is determined as the target object with the current moment.
  • the target object that matches this target object For example, if among the remaining target objects O 3 and O 4 at the previous moment, O 4 is the closest to the target object at the current moment, then the target object O 4 at the previous moment is determined to match the target object at the current moment. the target object, and the identification information of the target object at the current moment is determined as O 4 .
  • the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, or the target object is at the current moment and the previous moment.
  • the target object with the closest distance to the target object at the current moment is determined as the target object at the current moment.
  • the matching target object, and the identification information of the matching target object is used as the identification information of the target object, thereby adopting the distance-based greedy algorithm to determine the identification information of the target object, regardless of the visual feature of the target object, Thereby, the amount of calculation can be reduced, the time overhead can be reduced, and the real-time requirements of target tracking can be met.
  • the target tracking method provided by the embodiment of the present disclosure may be implemented by using a target tracking model.
  • the target tracking model may adopt Faster-RCNN (Faster Recurrent Convolutional Neural Network, faster recurrent convolutional neural network) or Fast-RCNN (Faster Recurrent Convolutional Neural Network, fast recursive convolutional neural network) or the like.
  • the backbone network of the target tracking model can adopt structures such as ResNet-18.
  • the target tracking model can be compressed by channel pruning, and at the same time, the multi-layer feature detection method is used to improve the model accuracy, so that the model accuracy is not reduced at the same time. , to increase model speed.
  • the method may further include: outputting the correspondence between the identification information of the target object and the coordinates.
  • the correspondence between the identification information and the coordinates may be output in the order of the identification information. For example, if the identification information includes the jersey numbers 1-11, the coordinates of the corresponding target objects can be output according to the order of the jersey numbers from small to large.
  • the motion trajectory of the target object may be obtained according to the coordinates of the target object at multiple times (for example, multiple consecutive times).
  • a Kalman filter algorithm may also be used to process the motion trajectory of the target object, so as to make the motion trajectory of the target object smoother.
  • the writing order of each step does not mean a strict execution order but constitutes any limitation on the implementation process, and the specific execution order of each step should be based on its function and possible Internal logic is determined.
  • the present disclosure also provides target tracking devices, electronic devices, computer-readable storage media, and programs, all of which can be used to implement any target tracking method provided by the present disclosure.
  • target tracking devices electronic devices, computer-readable storage media, and programs, all of which can be used to implement any target tracking method provided by the present disclosure.
  • FIG. 6 shows a block diagram of a target tracking apparatus provided by an embodiment of the present disclosure.
  • the target tracking device includes:
  • the first determining part 61 is configured to determine a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area;
  • the extraction part 62 is configured to extract the current feature of the target object in response to the situation that any target object in the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects;
  • the second determining part 63 is configured to determine, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object, wherein the plurality of reference features are in the plurality of target objects. Before the target objects collide with each other, the features extracted from the multiple target objects respectively;
  • the third determining part 64 is configured to determine, as the identification information of the target object, the identification information corresponding to the reference feature matching the current feature of the target object.
  • the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold
  • the extraction section 62 is configured to:
  • the apparatus further includes:
  • the third determination part 64 is configured to, for any target object, determine the first coordinates of the target object corresponding to the plurality of cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the acquisition according to the camera.
  • the first coordinate of the target object obtained from the image of ;
  • a fourth determining part configured to determine a sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras, wherein the tracking area includes a plurality of sub-areas ;
  • the fusion part is configured to fuse the first coordinates of the target object corresponding to the plurality of cameras according to the confidence of the plurality of cameras for the sub-region where the target object is located to obtain the fusion coordinates of the target object ;
  • the fifth determination part is configured to determine the distance between the target objects in the tracking area according to the fusion coordinates of the target objects in the tracking area.
  • the extraction part 62 is configured as:
  • Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
  • the extraction part 62 is configured as:
  • the multiple cameras determine the The camera that extracts the current features of this target object.
  • the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
  • the extraction section 62 is configured to:
  • the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition
  • the intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
  • the apparatus further includes:
  • the sixth determining part is configured to, for any one of the plurality of cameras, determine the confidence of the camera for the plurality of sub-areas according to the distance between the camera and the plurality of sub-areas of the tracking area.
  • the apparatus further includes:
  • the adjusting part is configured to, for any one of the plurality of cameras, adjust the confidence of the camera for the plurality of sub-regions according to the average distance between the target objects in the video frames collected by the camera.
  • the second determining part 63 is configured as:
  • the apparatus further includes:
  • a seventh determination part configured to respond to a situation where the similarity between the reference feature with the greatest similarity and the current feature of the target object is less than the similarity threshold, based on not conflicting with the target object and with any other target
  • the reference feature corresponding to the target object of the object conflict is determined, and the reference feature matching the current feature of the target object is determined.
  • the apparatus further includes:
  • the eighth determination part is configured to, for any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, the last moment does not conflict with other target objects.
  • the target object that conflicts with the target object and does not match with other target objects at the current moment is determined as the remaining target object at the previous moment; or, in response to the target object at the current moment and the previous moment In the case of conflict with the same other target object, the target object in the conflict zone to which the target object belongs at the last moment and which does not match with other target objects at the current moment is determined as the remaining target object at the last moment. target;
  • the ninth determination part is configured to, from the remaining target objects at the last moment, determine the target object with the closest distance to the target object at the current moment as the target matching the target object at the current moment object, and use the identification information of the matched target object as the identification information of the target object.
  • a "part" may be a part of a circuit, a part of a processor, a part of a program or software, etc., of course, a unit, a module or a non-modularity.
  • the present disclosure by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features.
  • the identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
  • the functions or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the methods described in the above method embodiments, and the implementation of the above method embodiments may refer to the descriptions of the above method embodiments. Repeat.
  • Embodiments of the present disclosure further provide a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium, or may be a volatile computer-readable storage medium.
  • Embodiments of the present disclosure also provide a computer program product, including computer-readable codes.
  • a processor in the device executes a method for implementing the target tracking method provided in any of the above embodiments. instruction.
  • Embodiments of the present disclosure further provide another computer program product for storing computer-readable instructions, which, when executed, cause the computer to perform the operations of the target tracking method provided by any of the foregoing embodiments.
  • Embodiments of the present disclosure further provide an electronic device, including: one or more processors; a memory for storing executable instructions; wherein the one or more processors are configured to invoke executable instructions stored in the memory instruction to execute the above method.
  • the electronic device may be provided as a terminal, server or other form of device.
  • FIG. 7 shows a block diagram of an electronic device 800 provided by an embodiment of the present disclosure.
  • electronic device 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, etc. terminal.
  • an electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, a sensor component 814 , and the communication component 816 .
  • the processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 802 can include one or more processors 820 to execute instructions to perform all or some of the steps of the methods described above.
  • processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components.
  • processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.
  • Memory 804 is configured to store various types of data to support operation at electronic device 800 . Examples of such data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, and the like. Memory 804 may be implemented by any type of volatile or nonvolatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Magnetic Disk
  • Power supply assembly 806 provides power to various components of electronic device 800 .
  • Power supply components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800 .
  • Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action.
  • the multimedia component 808 includes a front-facing camera and/or a rear-facing camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.
  • Audio component 810 is configured to output and/or input audio signals.
  • audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when electronic device 800 is in operating modes, such as calling mode, recording mode, and voice recognition mode.
  • the received audio signal may be further stored in memory 804 or transmitted via communication component 816 .
  • audio component 810 also includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.
  • Sensor assembly 814 includes one or more sensors for providing status assessment of various aspects of electronic device 800 .
  • the sensor assembly 814 can detect the on/off state of the electronic device 800, the relative positioning of the components, such as the display and the keypad of the electronic device 800, the sensor assembly 814 can also detect the electronic device 800 or one of the electronic device 800 Changes in the position of components, presence or absence of user contact with the electronic device 800 , orientation or acceleration/deceleration of the electronic device 800 and changes in the temperature of the electronic device 800 .
  • Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices.
  • the electronic device 800 may access wireless networks based on communication standards, such as Wi-Fi, 2G, 3G, 4G/LTE, 5G, or a combination thereof.
  • the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • electronic device 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A programmed gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation is used to perform the above method.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGA field programmable A programmed gate array
  • controller microcontroller, microprocessor or other electronic component implementation is used to perform the above method.
  • a non-volatile computer-readable storage medium such as a memory 804 comprising computer program instructions executable by the processor 820 of the electronic device 800 to perform the above method is also provided.
  • FIG. 8 shows a block diagram of an electronic device 1900 provided by an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server.
  • electronic device 1900 includes processing component 1922, which further includes one or more processors, and a memory resource represented by memory 1932 for storing instructions executable by processing component 1922, such as applications.
  • An application program stored in memory 1932 may include one or more modules, each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above-described methods.
  • the electronic device 1900 may also include a power supply assembly 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input output (I/O) interface 1958 .
  • Electronic device 1900 may operate based on an operating system stored in memory 1932, such as Windows Mac OS or similar.
  • a non-volatile computer-readable storage medium such as memory 1932 comprising computer program instructions executable by processing component 1922 of electronic device 1900 to perform the above-described method.
  • the present disclosure may be a system, method and/or computer program product.
  • the computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present disclosure.
  • a computer-readable storage medium may be a tangible device that can hold and store instructions for use by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Non-exhaustive list of computer readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory), static random access memory (SRAM), portable compact disk read only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically coded devices, such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • RAM random access memory
  • ROM read only memory
  • EPROM erasable programmable read only memory
  • flash memory static random access memory
  • SRAM static random access memory
  • CD-ROM compact disk read only memory
  • DVD digital versatile disk
  • memory sticks floppy disks
  • mechanically coded devices such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • Computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, light pulses through fiber optic cables), or through electrical wires transmitted electrical signals.
  • the computer readable program instructions described herein may be downloaded to various computing/processing devices from a computer readable storage medium, or to an external computer or external storage device over a network such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or network interface in each computing/processing device receives computer-readable program instructions from a network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • Computer program instructions for carrying out operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or instructions in one or more programming languages.
  • Source or object code written in any combination, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as the "C" language or similar programming languages.
  • the computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider through the Internet connect).
  • LAN local area network
  • WAN wide area network
  • custom electronic circuits such as programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs) can be personalized by utilizing state information of computer readable program instructions.
  • Computer readable program instructions are executed to implement various aspects of the present disclosure.
  • These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processor of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium on which the instructions are stored includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • Computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more functions for implementing the specified logical function(s) executable instructions.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.
  • the computer program product can be specifically implemented by hardware, software or a combination thereof.
  • the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), etc. Wait.
  • a software development kit Software Development Kit, SDK
  • the current feature of the target object is extracted, and based on the multiple reference features corresponding to the multiple target objects, the reference feature matching the current feature of the target object is determined, and the reference feature matching the current feature of the target object is determined.
  • the identification information corresponding to the reference feature of the feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained, Thus, the accuracy of multi-target tracking can be improved.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Signal Processing For Digital Recording And Reproducing (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

The present disclosure relates to an object tracking method and apparatus, an electronic device, and a storage medium. The method comprises: determining, according to distances between target objects in a tracking area, a plurality of target objects that conflict with each other; in response to a situation where any one of the plurality of target objects no longer conflicts with the other target objects in the plurality of target objects, extracting the current feature of the target object; determining, on the basis of a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object, the plurality of reference features being features extracted respectively for the plurality of target objects before the plurality of target objects conflict with each other; and determining identification information corresponding to the reference feature matching the current feature of the target object as identification information of the target object.

Description

目标跟踪方法及装置、电子设备和存储介质Target tracking method and device, electronic device and storage medium
相关申请的交叉引用CROSS-REFERENCE TO RELATED APPLICATIONS
本公开基于申请号为202011256801.6、申请日为2020年11月11日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本公开作为参考。The present disclosure is based on a Chinese patent application with application number 202011256801.6 and an application date of November 11, 2020, and claims the priority of the Chinese patent application, the entire contents of which are incorporated herein by reference.
技术领域technical field
本公开涉及计算机视觉技术领域，尤其涉及一种目标跟踪方法及装置、电子设备和存储介质。The present disclosure relates to the technical field of computer vision, and in particular, to a target tracking method and device, an electronic device and a storage medium.
背景技术Background technique
目标跟踪是指在连续的视频帧中跟踪同一目标对象(例如行人、车辆等)。随着计算机视觉技术的发展，目标跟踪算法在安防等领域应用广泛，在打造智慧生活方面具有重要价值。但在目标对象较密集的情况下，在目标跟踪的过程中，目标对象的标识(Identity document，ID)的一致性较差。Object tracking refers to tracking the same target object (eg pedestrian, vehicle, etc.) in consecutive video frames. With the development of computer vision technology, target tracking algorithms are widely used in security and other fields, and are of great value in building a smart life. However, in the case of dense target objects, in the process of target tracking, the consistency of the identity document (ID) of the target object is poor.
发明内容SUMMARY OF THE INVENTION
本公开提供了一种目标跟踪技术方案。The present disclosure provides a technical solution for target tracking.
根据本公开的一方面，提供了一种目标跟踪方法，包括：According to an aspect of the present disclosure, there is provided a target tracking method, comprising:
根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象；According to the distance between the target objects in the tracking area, determine multiple target objects that conflict with each other;
响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征；In response to a situation in which any one of the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects, extracting the current feature of the target object;
基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目标对象提取的特征；Based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is determined, wherein the plurality of reference features are the corresponding reference features of the target objects before the plurality of target objects collide with each other. Describe the features extracted from multiple target objects;
将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。The identification information corresponding to the reference feature matching the current feature of the target object is determined as the identification information of the target object.
在本公开实施例中，通过根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，并将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息，由此能够保持目标对象的标识信息的一致性，即，可以保持同一目标对象在进出冲突区时标识信息的一致性，从而能够提高多目标跟踪的准确性。In the embodiment of the present disclosure, by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features. The identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
在一种可能的实现方式中，彼此冲突的多个目标对象中的任一目标对象，与所述多个目标对象中的其他至少一个目标对象之间的距离小于或等于距离阈值；In a possible implementation manner, the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold;
所述响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，包括：The extracting the current feature of the target object in response to the situation that any target object in the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects includes:
对于所述多个目标对象中的任一目标对象，响应于该目标对象与所述多个目标对象中的其他目标对象之间的距离均大于所述距离阈值的情况，提取该目标对象的当前特征。For any target object in the plurality of target objects, in response to the situation that the distance between the target object and other target objects in the plurality of target objects is greater than the distance threshold, extract the current value of the target object feature.
根据该实现方式确定目标对象之间的冲突关系，并基于此进行目标跟踪，有助于提高目标跟踪的准确性和效率。According to this implementation, the conflict relationship between target objects is determined, and target tracking is performed based on this, which helps to improve the accuracy and efficiency of target tracking.
在一种可能的实现方式中，所述方法还包括：In a possible implementation, the method further includes:
对于任一目标对象，确定该目标对象对应于多个摄像头的第一坐标；其中，该目标对象对应于任一摄像头的第一坐标用于表征根据该摄像头采集的图像得到的该目标对象的第一坐标；For any target object, determine the first coordinates of the target object corresponding to multiple cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the first coordinates of the target object obtained according to the images collected by the camera. a coordinate;
根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，其中，所述跟踪区域包括多个子区域；According to the first coordinates of the target object corresponding to the plurality of cameras, determine the sub-area where the target object is located in the tracking area, wherein the tracking area includes a plurality of sub-areas;
根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头 的第一坐标进行融合，得到该目标对象的融合坐标；According to the confidence of the multiple cameras for the sub-region where the target object is located, the first coordinates of the target object corresponding to the multiple cameras are fused to obtain the fusion coordinates of the target object;
在所述根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象之前，还包括：Before determining the multiple conflicting target objects according to the distances between the target objects in the tracking area, the method further includes:
根据所述跟踪区域中目标对象的融合坐标，确定所述跟踪区域中目标对象之间的距离。The distance between the target objects in the tracking area is determined according to the fusion coordinates of the target objects in the tracking area.
在该实现方式中，通过结合目标对象对应于所述多个摄像头的第一坐标，得到该目标对象的融合坐标，并基于该目标对象的融合坐标进行目标跟踪，由此能够得到基于更精准的坐标进行目标跟踪，从而有助于提高目标跟踪的准确性。In this implementation, by combining the first coordinates of the target object corresponding to the plurality of cameras, the fusion coordinates of the target object are obtained, and the target tracking is performed based on the fusion coordinates of the target object, so that a more accurate The coordinates are used for target tracking, which helps to improve the accuracy of target tracking.
在一种可能的实现方式中，所述提取该目标对象的当前特征，包括：In a possible implementation manner, the extracting the current feature of the target object includes:
确定该目标对象在所述跟踪区域中所处的子区域；determining the sub-area where the target object is located in the tracking area;
根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头；Determine the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located;
根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征。Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
在该实现方式中，通过根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该实现方式提取的该目标对象的当前特征进行目标匹配，能够提高目标匹配的准确性。In this implementation manner, the camera for extracting the current feature of the target object is determined according to the confidence of the plurality of cameras for the sub-region where the target object is located, and the camera for extracting the current feature of the target object is determined according to the camera The current feature of the target object is extracted from the collected video frame, and the current feature of the target object thus extracted can have richer visual information. Therefore, performing target matching according to the current feature of the target object extracted by this implementation method can Improve the accuracy of target matching.
在一种可能的实现方式中，所述根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头，包括：In a possible implementation manner, determining the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located, includes:
根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头。According to the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the The camera that extracts the current features of this target object.
在该实现方式中，通过根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该实现方式提取的该目标对象的当前特征进行目标匹配，能够进一步提高目标匹配的准确性。In this implementation manner, according to the confidence of the sub-region where the target object is located by the plurality of cameras, and the difference between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the plurality of cameras The overlapping information of the target object is determined, the camera used for extracting the current feature of the target object is determined, and the current feature of the target object is extracted according to the video frame collected by the camera used for extracting the current feature of the target object, and the target object extracted from this The current feature of the target object can have richer visual information, therefore, target matching is performed according to the current feature of the target object extracted in this implementation manner, which can further improve the accuracy of target matching.
在一种可能的实现方式中，所述重叠信息包括该目标对象的检测框与其他目标对象的检测框的交并比；In a possible implementation manner, the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
所述根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头，包括：The determination is based on the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras. The camera used to extract the current features of the target object, including:
从满足重叠条件的摄像头中，将针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，其中，所述重叠条件表示在所述当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值。From the cameras that satisfy the overlapping condition, the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition The intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
在该实现方式中，通过将所述多个摄像头中当前采集的视频帧满足重叠条件的摄像头中，针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该实现方式提取的该目标对象的当前特征进行目标匹配，能够进一步提高目标匹配的准确性。In this implementation, the camera with the highest confidence in the sub-region where the target object is located is determined as the camera used for extracting the target object among the cameras whose currently collected video frames satisfy the overlapping condition. The current feature of the camera, and the current feature of the target object is extracted according to the video frame collected by the camera used to extract the current feature of the target object. The current feature of the target object thus extracted can have richer visual information. Therefore, , and perform target matching according to the current feature of the target object extracted by this implementation, which can further improve the accuracy of target matching.
在一种可能的实现方式中，所述方法还包括：In a possible implementation, the method further includes:
对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度。For any camera in the plurality of cameras, according to the distance between the camera and the plurality of sub-areas of the tracking area, the confidence level of the camera for the plurality of sub-areas is determined.
在该实现方式中对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度，并基于由此确定的置信度进行坐标融合和/或确定用于提取目标对象的当前特征的摄像头，从而有助于提高目标跟踪的准确性。In this implementation, for any camera in the plurality of cameras, the confidence of the camera for the plurality of sub-regions is determined according to the distance between the camera and the plurality of sub-regions of the tracking area, and based on the This determined confidence level performs coordinate fusion and/or determines the camera used to extract the current features of the target object, thereby helping to improve the accuracy of target tracking.
在一种可能的实现方式中，在所述确定所述多个摄像头针对所述多个子区域的置信度之后，所述方法还包括：In a possible implementation manner, after the determining the confidence levels of the plurality of cameras with respect to the plurality of sub-regions, the method further includes:
对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，调整该摄像头针对所述多个子区域的置信度。For any camera in the plurality of cameras, according to the average distance between the target objects in the video frames collected by the camera, the confidence level of the camera with respect to the plurality of sub-regions is adjusted.
在该实现方式中，通过对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，动态调整该摄像头针对所述多个子区域的置信度，并基于由此确定的置信度进行坐标融合和/或确定用于提取目标对象的当前特征的摄像头，从而有助于提高目标跟踪的准确性。In this implementation, for any camera in the plurality of cameras, according to the average distance between the target objects in the video frames collected by the camera, dynamically adjust the confidence level of the camera with respect to the plurality of sub-regions, and Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
在一种可能的实现方式中，所述基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，包括：In a possible implementation manner, the determining a reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects includes:
从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征；From the plurality of reference features corresponding to the plurality of target objects, determine the reference feature with the greatest similarity to the current feature of the target object;
响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。In response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold, determine the reference feature with the greatest similarity as the reference matching the current feature of the target object feature.
在该实现方式中，通过从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征，并响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征，由此在该目标对象与多个目标对象中的其他目标对象不再冲突的情况，首先基于所述多个目标对象对应的多个基准特征查找与该目标对象的当前特征匹配的基准特征，从而有助于提高查找与该目标对象的当前特征匹配的基准特征的速度和准确性。In this implementation manner, a reference feature with the greatest similarity to the current feature of the target object is determined from a plurality of reference features corresponding to the plurality of target objects, and the reference feature with the highest similarity is responsive to the reference feature with the highest similarity with the target object. In the case where the similarity between the current features of the target object is greater than or equal to the similarity threshold, the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object, so that the target object and multiple When other target objects in the target object are no longer in conflict, first, based on the plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is searched, thereby helping to improve the search efficiency of the target object. The speed and accuracy of the current feature matching the benchmark feature.
在一种可能的实现方式中，在所述确定与该目标对象的当前特征的相似度最大的基准特征之后，所述方法还包括：In a possible implementation manner, after the determining the reference feature with the greatest similarity to the current feature of the target object, the method further includes:
响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征。In response to the situation where the similarity between the reference feature with the greatest similarity and the current feature of the target object is less than the similarity threshold, based on the reference corresponding to the target object that does not conflict with the target object and conflicts with any other target object feature to determine the reference feature that matches the current feature of the target object.
在该实现方式中，通过响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征，由此能够在提高查找与该目标对象的当前特征匹配的基准特征的速度的前提下，进一步提高所确定的与该目标对象的当前特征匹配的基准特征的准确性。In this implementation, in response to the situation that the similarity between the reference feature with the highest similarity and the current feature of the target object is less than the similarity threshold, based on the fact that it does not conflict with the target object and is not in conflict with any other target object The reference feature corresponding to the conflicting target object is determined, and the reference feature matching the current feature of the target object is determined, so that the determined reference feature can be further improved on the premise of improving the speed of finding the reference feature matching the current feature of the target object. The accuracy of the fiducial feature matching the current features of this target object.
在一种可能的实现方式中，所述方法还包括：In a possible implementation, the method further includes:
对于任一目标对象，响应于该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突的情况，将所述上一时刻不与其他目标对象冲突、且未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；或者，响应于该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况，将该目标对象在所述上一时刻所属的冲突区中、未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；For any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, the last moment does not conflict with other target objects and does not conflict with all target objects. The target object matched by other target objects at the current moment is determined to be the remaining target objects at the previous moment; situation, the target object in the conflict zone to which the target object belongs at the last moment, and the target object that does not match with other target objects at the current moment, is determined as the remaining target object at the last moment;
从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息。From the remaining target objects at the last moment, the target object with the closest distance to the target object at the current moment is determined as the target object matching the target object at the current moment, and the matching target The identification information of the object is used as the identification information of the target object.
根据该实现方式能够减少计算量，降低时间开销，满足目标跟踪的实时性要求。According to this implementation manner, the amount of calculation and the time overhead can be reduced, and the real-time requirement of target tracking can be met.
根据本公开的一方面，提供了一种目标跟踪装置，包括：According to an aspect of the present disclosure, a target tracking device is provided, comprising:
第一确定部分，被配置为根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象；a first determining part, configured to determine a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area;
提取部分，被配置为响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征；an extraction part, configured to extract the current feature of the target object in response to a situation in which any one of the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects;
第二确定部分，被配置为基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目 标对象提取的特征；The second determination part is configured to determine, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature that matches the current feature of the target object, wherein the plurality of reference features are in the plurality of targets Before the objects collide with each other, the features extracted from the multiple target objects respectively;
第三确定部分，被配置为将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。The third determining part is configured to determine, as the identification information of the target object, the identification information corresponding to the reference feature matched with the current feature of the target object.
根据本公开的一方面，提供了一种电子设备，包括：一个或多个处理器；用于存储可执行指令的存储器；其中，所述一个或多个处理器被配置为调用所述存储器存储的可执行指令，以执行上述方法。According to an aspect of the present disclosure, there is provided an electronic device comprising: one or more processors; a memory for storing executable instructions; wherein the one or more processors are configured to invoke the memory storage executable instructions to perform the above method.
根据本公开的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。According to an aspect of the present disclosure, there is provided a computer-readable storage medium having computer program instructions stored thereon, the computer program instructions implementing the above method when executed by a processor.
根据本公开的一方面，提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行时实现上述方法。According to an aspect of the present disclosure, there is provided a computer program including computer readable code, which when executed in an electronic device, implements the above method when executed by a processor in the electronic device.
在本公开实施例中，通过根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，并将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息，由此能够保持目标对象的标识信息的一致性，即，可以保持同一目标对象在进出冲突区时标识信息的一致性，从而能够提高多目标跟踪的准确性。In the embodiment of the present disclosure, by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features. The identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.
根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings.
附图说明Description of drawings
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。The accompanying drawings, which are incorporated into and constitute a part of this specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the technical solutions of the present disclosure.
图1示出本公开实施例提供的目标跟踪方法的流程图。FIG. 1 shows a flowchart of a target tracking method provided by an embodiment of the present disclosure.
图2示出在第t时刻中彼此冲突的多个目标对象的示意图。FIG. 2 shows a schematic diagram of a plurality of target objects colliding with each other at time t.
图3示出在第t+1时刻中彼此冲突的多个目标对象的示意图。FIG. 3 shows a schematic diagram of a plurality of target objects colliding with each other at time t+1.
图4示出本公开实施例中多个目标对象以及多个目标对象对应的多个基准特征的示意图。FIG. 4 shows a schematic diagram of multiple target objects and multiple reference features corresponding to the multiple target objects in an embodiment of the present disclosure.
图5示出在本公开实施例中在跟踪区域中放置标志物的示意图。FIG. 5 shows a schematic diagram of placing markers in a tracking area in an embodiment of the present disclosure.
图6示出本公开实施例提供的目标跟踪装置的框图。FIG. 6 shows a block diagram of a target tracking apparatus provided by an embodiment of the present disclosure.
图7示出本公开实施例提供的一种电子设备800的框图。FIG. 7 shows a block diagram of an electronic device 800 provided by an embodiment of the present disclosure.
图8示出本公开实施例提供的一种电子设备1900的框图。FIG. 8 shows a block diagram of an electronic device 1900 provided by an embodiment of the present disclosure.
具体实施方式Detailed ways
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numbers in the figures denote elements that have the same or similar functions. While various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。The term "and/or" in this article is only an association relationship to describe the associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, it can mean that A exists alone, A and B exist at the same time, and A and B exist independently B these three cases. In addition, the term "at least one" herein refers to any combination of any one of the plurality or at least two of the plurality, for example, including at least one of A, B, and C, and may mean including from A, B, and C. Any one or more elements selected from the set of B and C.
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。In addition, in order to better illustrate the present disclosure, numerous specific details are set forth in the following detailed description. It will be understood by those skilled in the art that the present disclosure may be practiced without certain specific details. In some instances, methods, means, components and circuits well known to those skilled in the art have not been described in detail so as not to obscure the subject matter of the present disclosure.
图1示出本公开实施例提供的目标跟踪方法的流程图。所述目标跟踪方法的执行主体可以是目标跟踪装置。例如，所述目标跟踪方法可以由终端设备或服务器或其它处理设备执行。其中，终端设备可以是用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字助理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备或者可穿戴设备等。在一些可能 的实现方式中，所述目标跟踪方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。如图1所示，所述目标跟踪方法包括步骤S11至步骤S14。FIG. 1 shows a flowchart of a target tracking method provided by an embodiment of the present disclosure. The execution subject of the target tracking method may be a target tracking device. For example, the target tracking method may be performed by a terminal device or a server or other processing device. The terminal device may be a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, or a wearable devices, etc. In some possible implementations, the target tracking method may be implemented by a processor invoking computer-readable instructions stored in a memory. As shown in FIG. 1 , the target tracking method includes steps S11 to S14.
在步骤S11中，根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象。In step S11, according to the distance between the target objects in the tracking area, a plurality of target objects that conflict with each other are determined.
本公开实施例可以应用于智能监控、智能场景分析、安防、目标检测、目标跟踪等应用场景中。在本公开实施例中，跟踪区域表示需要进行目标跟踪的区域。本公开实施例中的跟踪区域可以是相对封闭的区域，也可以是相对开阔的区域。例如，跟踪区域可以是球场、商场、教室等。目标对象表示需要跟踪的对象。例如，目标对象可以是行人、车辆、运动场中的运动员(例如足球场中的球员)等任何需要跟踪的对象。在本公开实施例中，彼此冲突的目标对象可以表示彼此之间距离较近的目标对象。The embodiments of the present disclosure can be applied to application scenarios such as intelligent monitoring, intelligent scene analysis, security protection, target detection, and target tracking. In the embodiment of the present disclosure, the tracking area represents an area where target tracking needs to be performed. The tracking area in the embodiment of the present disclosure may be a relatively closed area or a relatively open area. For example, the tracking area may be a stadium, a mall, a classroom, and the like. The target object represents the object that needs to be tracked. For example, the target object can be any object that needs to be tracked, such as a pedestrian, a vehicle, an athlete in a sports field (eg, a player in a football field). In an embodiment of the present disclosure, target objects that conflict with each other may represent target objects that are closer to each other.
在本公开实施例中，可以针对摄像头采集的每一视频帧，分别确定跟踪区域中目标对象之间的距离，从而确定在该视频帧对应的时刻目标对象之间彼此冲突的情况。当然，也可以不对摄像头采集的视频帧逐帧进行分析，例如，可以每隔若干个视频帧，确定跟踪区域中目标对象之间的距离，从而确定在该视频帧对应的时刻目标对象之间彼此冲突的情况。In the embodiment of the present disclosure, for each video frame collected by the camera, the distance between the target objects in the tracking area can be determined respectively, so as to determine the conflict between the target objects at the moment corresponding to the video frame. Of course, the video frames collected by the camera may not be analyzed frame by frame. For example, the distance between the target objects in the tracking area may be determined every several video frames, so as to determine the distance between the target objects at the moment corresponding to the video frame. conflict situation.
在本公开实施例中，在任一时刻中，可能存在一组或多组彼此冲突的目标对象，当然，也可能不存在彼此冲突的目标对象。图2示出在第t时刻中彼此冲突的多个目标对象的示意图。如图2所示，在第t时刻，目标对象O 1、O 4与O 3彼此冲突，目标对象O 2与O 5彼此冲突，目标对象O 6与O 7彼此冲突。 In the embodiment of the present disclosure, at any moment, there may be one or more groups of conflicting target objects, and of course, there may be no conflicting target objects. FIG. 2 shows a schematic diagram of a plurality of target objects colliding with each other at time t. As shown in FIG. 2 , at time t, the target objects O 1 , O 4 and O 3 conflict with each other, the target objects O 2 and O 5 conflict with each other, and the target objects O 6 and O 7 conflict with each other.
在一种可能的实现方式中，可以将彼此冲突的多个目标对象加入同一个冲突区中，即，可以认为彼此冲突的多个目标对象处于同一个冲突区中。其中，冲突区可以是虚拟的区域，任一冲突区中的目标对象之间彼此冲突。例如，在图2中，在第t时刻，冲突区1中包括目标对象O 1、O 4和O 3，冲突区2中包括目标对象O 2和O 5，冲突区3中包括目标对象O 6和O 7。 In a possible implementation manner, multiple target objects that conflict with each other may be added to the same conflict area, that is, multiple target objects that conflict with each other may be considered to be in the same conflict area. The conflict area may be a virtual area, and the target objects in any conflict area conflict with each other. For example, in FIG. 2 , at time t, the conflict zone 1 includes target objects O 1 , O 4 and O 3 , the conflict zone 2 includes target objects O 2 and O 5 , and the conflict zone 3 includes target objects O 6 and O 7 .
在步骤S12中，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征。In step S12, in response to the situation that any one of the multiple target objects is no longer in conflict with other target objects of the multiple target objects, the current feature of the target object is extracted.
图3示出在第t+1时刻中彼此冲突的多个目标对象的示意图。如图3所示，在第t+1时刻，冲突区1中包括目标对象O 1和O 4，冲突区2中包括目标对象O 2和O 5，冲突区3中包括目标对象O 6和O 7。即，在第t+1时刻，目标对象O 1与O 4彼此冲突，目标对象O 2与O 5彼此冲突，目标对象O 6与O 7彼此冲突。比较第t+1时刻与第t时刻可知，在第t+1时刻，目标对象O 3与O 1、O 4不再冲突；也就是说，场景中的所有目标对象在任意时刻都动态的聚类成多个冲突区。 FIG. 3 shows a schematic diagram of a plurality of target objects colliding with each other at time t+1. As shown in FIG. 3 , at time t+1, conflict area 1 includes target objects O 1 and O 4 , conflict area 2 includes target objects O 2 and O 5 , and conflict area 3 includes target objects O 6 and O 7 . That is, at time t+1, the target objects O 1 and O 4 conflict with each other, the target objects O 2 and O 5 conflict with each other, and the target objects O 6 and O 7 conflict with each other. Comparing the time t+1 and the time t, we can see that at the time t+1, the target object O 3 is no longer in conflict with O 1 and O 4 ; that is, all the target objects in the scene converge dynamically at any time. Classes into multiple conflict zones.
在一种可能的实现方式中，相邻的两个时刻之间的时间间隔可以等于摄像头采集视频帧的帧率的倒数。在另一种可能的实现方式中，相邻的两个时刻之间的时间间隔可以大于摄像头采集视频帧的帧率的倒数，例如可以等于摄像头采集视频帧的帧率的倒数的H倍，其中，H是大于1的整数。In a possible implementation manner, the time interval between two adjacent moments may be equal to the inverse of the frame rate at which the video frame is captured by the camera. In another possible implementation manner, the time interval between two adjacent moments may be greater than the inverse of the frame rate at which the camera captures video frames, for example, it may be equal to H times the inverse of the frame rate at which the camera captures video frames, wherein , H is an integer greater than 1.
在一种可能的实现方式中，彼此冲突的多个目标对象中的任一目标对象，与所述多个目标对象中的其他至少一个目标对象之间的距离小于或等于距离阈值；所述响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，包括：对于所述多个目标对象中的任一目标对象，响应于该目标对象与所述多个目标对象中的其他目标对象之间的距离均大于所述距离阈值的情况，提取该目标对象的当前特征。In a possible implementation manner, the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold; the response In a situation where any one of the multiple target objects is no longer in conflict with other target objects in the multiple target objects, extracting the current feature of the target object includes: for any one of the multiple target objects The target object, in response to the situation that the distances between the target object and other target objects in the plurality of target objects are all greater than the distance threshold, extract the current feature of the target object.
在一些实施例中，可以采用具有噪声的基于密度聚类方法(Density-Based Spatial Clustering of Applications with Noise，DBSCAN)，将与其他目标对象之间的距离小于距离阈值的目标对象进行聚类，形成冲突区。In some embodiments, a noise-based density clustering method (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) can be used to cluster target objects whose distances from other target objects are less than a distance threshold to form conflict zone.
在该实现方式中，在任一帧中，可以计算任意两个目标对象之间的距离，若该两个目标对象之间的距离小于或等于距离阈值，则可以确定该两个目标对象彼此冲突。若存在另一个目标对象与该两个目标对象中的至少一个目标对象之间的距离小于或等于距离阈值，则可以确定这三个目标对象彼此冲突。例如，在第t时刻，若目标对象O 1与目标对象O 4之间的距离小于或等于距离阈值，则可以确定目标对象O 1与目标对象O 4此冲突。若目标对象O 3与目标对象O 4之间的距离小于或等于距离阈值，则可以确定目标对象O 1、O 4与O 3彼此冲突。 In this implementation, in any frame, the distance between any two target objects can be calculated, and if the distance between the two target objects is less than or equal to a distance threshold, it can be determined that the two target objects conflict with each other. If the distance between another target object and at least one of the two target objects is less than or equal to the distance threshold, it may be determined that the three target objects conflict with each other. For example, at time t, if the distance between the target object O1 and the target object O4 is less than or equal to the distance threshold, it can be determined that the target object O1 and the target object O4 conflict. If the distance between the target object O 3 and the target object O 4 is less than or equal to the distance threshold, it can be determined that the target objects O 1 , O 4 and O 3 conflict with each other.
作为该实现方式的一个示例，在判断目标对象之间是否彼此冲突时，可以仅考虑目标对象的位置(例如融合坐标)之间的距离是否小于或等于距离阈值，例如，若任意两个目标对象的位置之间的距离小于或等于距离阈值，则可以确定该两个目标对象彼此冲突。作为该实现方式的另一个示例，可以不仅考虑目标对象的位置之间的距离，还考虑目标对象的视觉特征(例如基准特征和/或当前特征)之间的距离，例如，若任意两个目标对象的位置之间的距离小于或等于第一距离阈值，且该两个目标对象的视觉特征之间的距离小于或等于第二距离阈值，则可以确定该两个目标对象彼此冲突。其中，目标对象的视觉特征之间的距离可以表示目标对象的视觉特征之间的相似度。As an example of this implementation, when judging whether the target objects conflict with each other, it may only consider whether the distance between the positions of the target objects (for example, fusion coordinates) is less than or equal to the distance threshold, for example, if any two target objects If the distance between the positions is less than or equal to the distance threshold, it can be determined that the two target objects collide with each other. As another example of this implementation, not only the distance between the positions of the target objects, but also the distance between the visual features (such as the reference feature and/or the current feature) of the target objects may be considered, for example, if any two targets If the distance between the positions of the objects is less than or equal to the first distance threshold, and the distance between the visual features of the two target objects is less than or equal to the second distance threshold, it can be determined that the two target objects conflict with each other. The distance between the visual features of the target object may represent the similarity between the visual features of the target object.
在该实现方式中，若某一目标对象与另一目标对象之间的距离小于或等于距离阈值，则可以确定该两个目标对象彼此冲突；若某一目标对象与其他目标对象的距离均大于所述距离阈值，则可以确定该目 标对象不与任何其他目标对象冲突。根据该实现方式确定目标对象之间的冲突关系，并基于此进行目标跟踪，有助于提高目标跟踪的准确性和效率。In this implementation, if the distance between a certain target object and another target object is less than or equal to the distance threshold, it can be determined that the two target objects conflict with each other; if the distance between a certain target object and other target objects is greater than the distance threshold, it can be determined that the target object does not conflict with any other target object. According to this implementation, the conflict relationship between target objects is determined, and target tracking is performed based on this, which helps to improve the accuracy and efficiency of target tracking.
在另一种可能的实现方式中，彼此冲突的多个目标对象中的任一目标对象，与所述多个目标对象中的其他所有目标对象之间的距离均小于或等于距离阈值；所述响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，包括：对于所述多个目标对象中的任一目标对象，当该目标对象与所述多个目标对象中的任一目标对象之间的距离大于所述距离阈值时，提取该目标对象的当前特征。In another possible implementation manner, the distance between any target object among the multiple target objects that conflict with each other and all other target objects among the multiple target objects is less than or equal to a distance threshold; the In response to a situation in which any one of the plurality of target objects no longer conflicts with other target objects in the plurality of target objects, extracting the current feature of the target object includes: for any one of the plurality of target objects; A target object, when the distance between the target object and any target object among the plurality of target objects is greater than the distance threshold, extract the current feature of the target object.
在一种可能的实现方式中，所述提取该目标对象的当前特征，包括：确定该目标对象在所述跟踪区域中所处的子区域；根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头；根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征。In a possible implementation manner, the extracting the current feature of the target object includes: determining a sub-area where the target object is located in the tracking area; Determine the camera used for extracting the current feature of the target object; according to the video frame collected by the camera used for extracting the current feature of the target object, extract the current feature of the target object.
在该实现方式中，所述跟踪区域可以包括至少一个子区域，例如，所述跟踪区域可以划分为多个子区域。在该实现方式中，通过采用多个摄像头进行目标跟踪，能够获得多个摄像头对应的多个视角的视觉信息，多个视角捕捉到的视觉信息可以互相补充，从而有助于提高目标检测和跟踪的准确性。In this implementation, the tracking area may include at least one sub-area, for example, the tracking area may be divided into a plurality of sub-areas. In this implementation, by using multiple cameras for target tracking, visual information from multiple viewing angles corresponding to multiple cameras can be obtained, and the visual information captured by multiple viewing angles can complement each other, thereby helping to improve target detection and tracking. accuracy.
在该实现方式中，通过根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该实现方式提取的该目标对象的当前特征进行目标匹配，能够提高目标匹配的准确性。In this implementation manner, the camera for extracting the current feature of the target object is determined according to the confidence of the plurality of cameras for the sub-region where the target object is located, and the camera for extracting the current feature of the target object is determined according to the camera The current feature of the target object is extracted from the collected video frame, and the current feature of the target object thus extracted can have richer visual information. Therefore, performing target matching according to the current feature of the target object extracted by this implementation method can Improve the accuracy of target matching.
作为该实现方式的一个示例，所述根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头，包括：根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头。As an example of the implementation manner, the determining a camera for extracting the current feature of the target object according to the confidence of the sub-region where the target object is located by the plurality of cameras includes: according to the plurality of cameras for the target object The confidence level of the sub-region where it is located, and the overlapping information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the method used to extract the current feature of the target object. Camera.
在该示例中，通过根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该示例提取的该目标对象的当前特征进行目标匹配，能够进一步提高目标匹配的准确性。In this example, according to the confidence of the sub-region where the target object is located by the plurality of cameras, and the difference between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the plurality of cameras Overlapping information, determine the camera used for extracting the current feature of the target object, and extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object, and extract the current feature of the target object. The current feature can have richer visual information, therefore, performing target matching according to the current feature of the target object extracted from this example can further improve the accuracy of target matching.
在一个例子中，所述重叠信息包括该目标对象的检测框与其他目标对象的检测框的交并比；所述根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头，包括：从满足重叠条件的摄像头中，将针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，其中，所述重叠条件表示在所述当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值。In one example, the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects; the confidence level of the sub-region where the target object is located according to the plurality of cameras, and the The overlapping information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by multiple cameras, and determining the camera used to extract the current feature of the target object, including: from the cameras that meet the overlapping conditions, The camera with the highest confidence for the sub-region where the target object is located is determined as the camera used for extracting the current feature of the target object, wherein the overlapping condition indicates that the target object is in the currently collected video frame. The intersection ratios of the detection frame and detection frames of other target objects are all smaller than a predetermined threshold.
例如，目标对象所处的子区域是子区域A 1，所述多个摄像头包括摄像头C 1、摄像头C 2、摄像头C 3和摄像头C 4，摄像头C 1针对子区域A 1的置信度是Z 11，摄像头C 2针对子区域A 1的置信度是Z 21，摄像头C 3针对子区域A 1的置信度是Z 31，摄像头C 4针对子区域A 1的置信度是Z 41，Z 11＞Z 21＞Z 31＞Z 41。若在摄像头C 1当前采集的视频帧中，不存在与该目标对象的检测框的交并比大于或等于预定阈值的其他检测框，即，在摄像头C 1当前采集的视频帧中，该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值，则根据摄像头C 1采集的视频帧，提取该目标对象的当前特征。若在摄像头C 1当前采集的视频帧中，存在与该目标对象的检测框的交并比大于或等于预定阈值的其他检测框，则判断摄像头C 2当前采集的视频帧中，是否存在与该目标对象的检测框的交并比大于或等于预定阈值的其他检测框，即，判断摄像头C 2当前采集的视频帧中，该目标对象的检测框与其他目标对象的检测框的交并比是否均小于预定阈值。若摄像头C 2当前采集的视频帧中，不存在与该目标对象的检测框的交并比大于或等于预定阈值的其他检测框，即，在摄像头C 2当前采集的视频帧中，该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值，则根据摄像头C 2采集的视频帧，提取该目标对象的当前特征，以此类推；也就是说，按照摄像头针对子区域的置信度排序，在置信度高的摄像头当前采集的视频帧中，该目标对象被遮挡的情况下，可以切换到置信度次之的摄像头，以此类推；使提取该目标对象的当前特征的摄像头在不被遮挡的情况下，尽可能为置信度高的摄像头，减小了漏检率，提高了目标匹配的准确性。 For example, the sub-area where the target object is located is the sub-area A 1 , the multiple cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4 , and the confidence level of the camera C 1 for the sub-area A 1 is Z 11. The confidence level of the camera C2 for the sub-region A1 is Z 21 , the confidence level of the camera C3 for the sub - region A 1 is Z 31 , the confidence level of the camera C 4 for the sub -region A 1 is Z 41 , Z 11 > Z 21 >Z 31 >Z 41 . If there is no other detection frame whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold in the video frame currently collected by the camera C1 , that is, in the video frame currently collected by the camera C1 , the target object If the intersection ratio between the detection frame of the object and the detection frames of other target objects is smaller than the predetermined threshold, the current feature of the target object is extracted according to the video frame collected by the camera C1 . If in the video frame currently collected by the camera C1 , there are other detection frames whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold, then it is judged whether there is a video frame currently collected by the camera C2 that is related to the target object. The intersection ratio of the detection frame of the target object is greater than or equal to other detection frames with a predetermined threshold, that is, to determine whether the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera C 2 is not. are less than the predetermined threshold. If there is no other detection frame whose intersection ratio with the detection frame of the target object is greater than or equal to a predetermined threshold in the video frame currently collected by the camera C2, that is, in the video frame currently collected by the camera C2 , the target object The intersection ratio of the detection frame of the target object and the detection frame of other target objects is smaller than the predetermined threshold, then according to the video frame collected by the camera C 2 , the current feature of the target object is extracted, and so on; In the video frame currently collected by the camera with high confidence, if the target object is occluded, you can switch to the camera with the next highest confidence, and so on; Under the condition that the camera is not blocked, it is a camera with high confidence as much as possible, which reduces the missed detection rate and improves the accuracy of target matching.
在这个例子中，对于所述多个摄像头中的任一摄像头，该摄像头当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比，可以表征该目标对象被其他目标对象遮挡的比例。该摄像头当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比越大，则可以表征该目标对象被其他目标对象遮挡的比例越大；该摄像头当前采集的视频帧中该目标对象的检测框与其他目标对象 的检测框的交并比越小，则可以表征该目标对象被其他目标对象遮挡的比例越小。在该摄像头当前采集的视频帧中，该目标对象被其他目标对象遮挡的比例越小，则基于该摄像头当前采集的视频帧所提取的该目标对象的当前特征所具备的该目标对象的视觉信息越丰富，例如，能够具备该目标对象的更多特征点的信息。这个例子通过将所述多个摄像头中当前采集的视频帧满足重叠条件的摄像头中，针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，并根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征，由此提取的该目标对象的当前特征能够具备更丰富的视觉信息，因此，根据该例子提取的该目标对象的当前特征进行目标匹配，能够进一步提高目标匹配的准确性。In this example, for any camera among the plurality of cameras, the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera can indicate that the target object is affected by other target objects. The proportion of target object occlusion. The larger the intersection ratio between the detection frame of the target object and the detection frames of other target objects in the video frame currently collected by the camera, the greater the proportion of the target object being occluded by other target objects; the video currently collected by the camera is larger. The smaller the intersection ratio of the detection frame of the target object and the detection frames of other target objects in the frame, the smaller the proportion of the target object being occluded by other target objects. In the video frame currently collected by the camera, the smaller the proportion that the target object is occluded by other target objects, the visual information of the target object possessed by the current features of the target object extracted based on the video frame currently collected by the camera. The richer, for example, the more feature point information of the target object can be provided. In this example, the camera with the highest confidence for the sub-region where the target object is located among the cameras whose currently collected video frames meet the overlapping condition, is determined as the camera for extracting the current feature of the target object , and extract the current feature of the target object according to the video frame collected by the camera used to extract the current feature of the target object, so that the current feature of the target object extracted can have richer visual information. Therefore, according to this example The extracted current features of the target object are used for target matching, which can further improve the accuracy of target matching.
在步骤S13中，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目标对象提取的特征。In step S13, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is determined, wherein the plurality of reference features are when the plurality of target objects conflict with each other Before, the features extracted for the multiple target objects are respectively.
图4示出本公开实施例中多个目标对象以及多个目标对象对应的多个基准特征的示意图。如图4所示，所述多个目标对象包括目标对象O 1、目标对象O 2、目标对象O 3和目标对象O 4，其中，目标对象O 1的基准特征是基准特征F 01，目标对象O 2的基准特征是基准特征F 02，目标对象O 3的基准特征是基准特征F 03，目标对象O 4的基准特征是基准特征F 04。 FIG. 4 shows a schematic diagram of multiple target objects and multiple reference features corresponding to the multiple target objects in an embodiment of the present disclosure. As shown in FIG. 4 , the plurality of target objects include target object O 1 , target object O 2 , target object O 3 and target object O 4 , wherein the reference feature of target object O 1 is reference feature F 01 , and the target object The reference feature of O 2 is the reference feature F 02 , the reference feature of the target object O 3 is the reference feature F 03 , and the reference feature of the target object O 4 is the reference feature F 04 .
在一种可能的实现方式中，可以在首帧提取目标对象的基准特征，并存储目标对象的基准特征。例如，可以在首帧提取所有目标对象的基准特征，并存储所有目标对象的基准特征。当然，若首帧中部分目标对象被严重遮挡，则可以在之后的视频帧中提取这些目标对象的基准特征，以获得质量更高的、包含更丰富的目标对象的视觉信息的基准特征。在提取了目标对象的基准特征之后，可以存储目标对象的标识信息与目标对象的基准特征之间的对应关系。在本公开实施例中，确定用于提取目标对象的基准特征的摄像头的方式，与上文中确定用于提取目标对象的当前特征的摄像头的方式类似，在此不再赘述。In a possible implementation manner, the reference feature of the target object may be extracted in the first frame, and the reference feature of the target object may be stored. For example, the fiducial features of all target objects can be extracted in the first frame, and the fiducial features of all target objects can be stored. Of course, if some of the target objects in the first frame are severely occluded, the reference features of these target objects can be extracted in subsequent video frames to obtain higher-quality reference features containing richer visual information of the target objects. After the reference feature of the target object is extracted, the correspondence between the identification information of the target object and the reference feature of the target object may be stored. In this embodiment of the present disclosure, the manner of determining the camera for extracting the reference feature of the target object is similar to the manner in which the camera for extracting the current feature of the target object is determined above, and details are not described herein again.
作为该实现方式的一个示例，在提取目标对象的基准特征后，若在后续目标跟踪的过程中，提取了该目标对象的当前特征，则可以根据所提取的该目标对象的当前特征，更新该目标对象的基准特征，以提高根据存储的基准特征进行目标匹配的准确性。例如，可以将存储的目标对象的基准特征与该目标对象的当前特征加权，得到新的基准特征。当然，也可以不更新目标对象的基准特征，以减少计算量。As an example of this implementation, after the reference feature of the target object is extracted, if the current feature of the target object is extracted in the subsequent target tracking process, the current feature of the target object can be updated according to the extracted current feature of the target object. The fiducial features of the target object to improve the accuracy of target matching based on the stored fiducial features. For example, the stored reference feature of the target object may be weighted with the current feature of the target object to obtain a new reference feature. Of course, the reference feature of the target object may not be updated to reduce the amount of computation.
在一种可能的实现方式中，在目标对象是人(例如行人、球场上的球员等)的情况下，可以通过ReID(person Re-IDentificaion，行人重识别)模块提取目标对象的特征。例如，可以通过ReID模块提取目标对象的基准特征和/或当前特征。在一个例子中，ReID模块可以采用卷积神经网络来实现。当然，也可以采用其他特征提取方法提取目标对象的特征，只要所提取的目标对象的特征能够体现目标对象的视觉信息即可。In a possible implementation manner, when the target object is a person (for example, a pedestrian, a player on a court, etc.), the feature of the target object can be extracted by a ReID (person Re-IDentificaion, pedestrian re-identification) module. For example, the reference feature and/or the current feature of the target object can be extracted by the ReID module. In one example, the ReID module can be implemented using a convolutional neural network. Of course, other feature extraction methods can also be used to extract the features of the target object, as long as the extracted features of the target object can reflect the visual information of the target object.
在一种可能的实现方式中，所述基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，包括：从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征；响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。其中，相似度可以是余弦相似度等。In a possible implementation manner, the determining, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature that matches the current feature of the target object includes: selecting from a plurality of reference features corresponding to the plurality of target objects Among the reference features, the reference feature with the greatest similarity with the current feature of the target object is determined; in response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold , and the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object. The similarity may be cosine similarity or the like.
例如，在第t时刻，冲突区1中包括目标对象O 1、O 4和O 3，冲突区1对应的基准特征包括目标对象O 1的基准特征F 01、目标对象O 4的基准特征F 04和目标对象O 3的基准特征F 03。在第t+1时刻，检测到有一个目标对象与其余目标对象的距离大于距离阈值的情况下，表示该目标对象脱离冲突区1，即，冲突区1中的一个目标对象与另外两个目标对象不再冲突，那么，提取该目标对象的当前特征F 1n，分别计算冲突区1对应的3个基准特征F 01、F 04和F 03与该目标对象的当前特征F 1n之间的相似度。若该3个基准特征中，与该目标对象的当前特征F 1n相似度最大的是基准特征F 03，且基准特征F 03与该目标对象的当前特征F 1n之间的相似度大于或等于相似度阈值，则可以将基准特征F 03确定为与该目标对象的当前特征匹配的基准特征。 For example, at time t, the conflict area 1 includes target objects O 1 , O 4 and O 3 , and the reference features corresponding to the conflict area 1 include the reference feature F 01 of the target object O 1 and the reference feature F 04 of the target object O 4 and the reference feature F 03 of the target object O 3 . At time t+1, when it is detected that the distance between one target object and the rest of the target objects is greater than the distance threshold, it means that the target object leaves the conflict zone 1, that is, one target object in the conflict zone 1 and the other two targets If the objects no longer conflict, then extract the current feature F 1n of the target object, and calculate the similarity between the three reference features F 01 , F 04 and F 03 corresponding to the conflict area 1 and the current feature F 1n of the target object respectively . If among the three reference features, the reference feature F 03 has the greatest similarity with the current feature F 1n of the target object, and the similarity between the reference feature F 03 and the current feature F 1n of the target object is greater than or equal to similarity If the degree threshold is set, the reference feature F 03 can be determined as the reference feature matching the current feature of the target object.
在该实现方式中，通过从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征，并响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征，由此在该目标对象与多个目标对象中的其他目标对象不再冲突的情况，首先基于所述多个目标对象对应的多个基准特征查找与该目标对象的当前特征匹配的基准特征，从而有助于提高查找与该目标对象的当前特征匹配的基准特征的速度和准确性。In this implementation manner, a reference feature with the greatest similarity to the current feature of the target object is determined from a plurality of reference features corresponding to the plurality of target objects, and the reference feature with the highest similarity is responsive to the reference feature with the highest similarity with the target object. In the case where the similarity between the current features of the target object is greater than or equal to the similarity threshold, the reference feature with the greatest similarity is determined as the reference feature matching the current feature of the target object, so that the target object and multiple When other target objects in the target object are no longer in conflict, first, based on the plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is searched, thereby helping to improve the search efficiency of the target object. The speed and accuracy of the current feature matching the benchmark feature.
作为该实现方式的一个示例，在所述确定与该目标对象的当前特征的相似度最大的基准特征之后，所述方法还包括：响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与 该目标对象的当前特征匹配的基准特征。As an example of this implementation, after the determining the reference feature with the greatest similarity to the current feature of the target object, the method further includes: in response to the reference feature with the greatest similarity and the current feature of the target object When the similarity between them is less than the similarity threshold, based on the reference feature corresponding to the target object that does not conflict with the target object and conflicts with any other target object, the reference feature matching the current feature of the target object is determined.
在该示例中，若在所述多个目标对象对应的多个基准特征中，找不到与该目标对象的当前特征匹配的基准特征，则可以在其他冲突区对应的基准特征(即，不与该目标对象冲突，但与其他目标对象冲突的目标对象对应的基准特征)中查找与该目标对象的当前特征匹配的基准特征。例如，若所述多个目标对象对应的多个基准特征与该目标对象的当前特征的相似度均小于相似度阈值，则可以确定在所述多个目标对象对应的多个基准特征中，找不到与该目标对象的当前特征匹配的基准特征。在该示例中，通过响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征，由此能够在提高查找与该目标对象的当前特征匹配的基准特征的速度的前提下，进一步提高所确定的与该目标对象的当前特征匹配的基准特征的准确性。In this example, if a reference feature matching the current feature of the target object cannot be found among the reference features corresponding to the target objects, the reference features corresponding to other conflict areas (that is, not The reference feature corresponding to the target object that conflicts with the target object but conflicts with other target objects) is searched for the reference feature that matches the current feature of the target object. For example, if the similarity between the multiple reference features corresponding to the multiple target objects and the current feature of the target object is less than the similarity threshold, it may be determined that among the multiple reference features corresponding to the multiple target objects, find the There are no fiducial features that match the current features of this target object. In this example, by responding to the situation that the similarity between the reference feature with the highest similarity and the current feature of the target object is less than the similarity threshold, based on not conflicting with the target object, and conflicting with any other target object The reference feature corresponding to the target object is determined, and the reference feature matching the current feature of the target object is determined, so as to improve the speed of finding the reference feature matching the current feature of the target object. The accuracy with which the current features of this target object match the benchmark features.
在一个例子中，若在所有冲突区中均找不到与该目标对象的当前特征匹配的基准特征，则可以在非冲突区中查找与该目标对象的当前特征匹配的基准特征。In one example, if a reference feature matching the current feature of the target object cannot be found in all conflict regions, a reference feature matching the current feature of the target object can be searched in the non-conflict region.
在另一种可能的实现方式中，所述基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，包括：确定所述多个目标对象对应的多个基准特征中，与该目标对象的当前特征的相似度最大的基准特征；将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。In another possible implementation manner, the determining, based on the plurality of reference features corresponding to the plurality of target objects, the reference features matching the current features of the target objects includes: determining the reference features corresponding to the plurality of target objects. Among the plurality of reference features, the reference feature with the greatest similarity with the current feature of the target object; the reference feature with the greatest similarity is determined as the reference feature matched with the current feature of the target object.
在步骤S14中，将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。In step S14, the identification information corresponding to the reference feature matching the current feature of the target object is determined as the identification information of the target object.
在本公开实施例中，目标对象的标识信息可以是能够用于唯一标识目标对象的信息，例如可以是ID、编号、姓名等。例如，若跟踪区域是足球场，目标对象包括足球场上的球员，则目标对象的标识信息可以是球员所属球队和球衣号码。In the embodiment of the present disclosure, the identification information of the target object may be information that can be used to uniquely identify the target object, for example, may be an ID, a serial number, a name, and the like. For example, if the tracking area is a football field and the target object includes players on the football field, the identification information of the target object may be the team to which the player belongs and the jersey number.
例如，与该目标对象的当前特征F 1n匹配的基准特征是F 03，基准特征F 03对应的标识信息是O 3，则可以将O 3确定为该目标对象的标识信息。 For example, if the reference feature matching the current feature F 1n of the target object is F 03 , and the identification information corresponding to the reference feature F 03 is O 3 , then O 3 may be determined as the identification information of the target object.
在一种可能的实现方式中，任一目标对象对应的目标跟踪结果还可以包括该目标对象的位置信息。利用本公开实施例提供的目标跟踪方法提供的目标跟踪结果，可以对跟踪区域中目标对象的行为进行分析。例如，跟踪区域是足球场，目标对象包括足球场上的球员，则根据球员的标识信息(例如所属球队和球衣号码)和球员在至少一个视频帧中的位置信息，可以对球员的行为进行分析，例如分析球员是否越位等。In a possible implementation manner, the target tracking result corresponding to any target object may also include position information of the target object. Using the target tracking result provided by the target tracking method provided by the embodiment of the present disclosure, the behavior of the target object in the tracking area can be analyzed. For example, if the tracking area is a football field, and the target object includes players on the football field, then according to the identification information of the players (such as the team and jersey numbers) and the position information of the players in at least one video frame, the behavior of the players can be determined. Analysis, such as analyzing whether the player is offside, etc.
在本公开实施例中，通过根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，并将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息，由此能够保持目标对象的标识信息的一致性，即，可以保持同一目标对象在进出冲突区时标识信息的一致性，从而能够提高多目标跟踪的准确性。本公开实施例能够适用于复杂的目标跟踪场景，例如可以适用于跟踪区域中目标对象的尺寸不一、跟踪区域中存在小尺寸难检测的目标对象、跟踪区域中目标对象较密集、目标对象的外观相近、目标对象的运动复杂度较高、存在严重遮挡等目标跟踪场景。In the embodiment of the present disclosure, by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features. The identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking. The embodiments of the present disclosure can be applied to complex target tracking scenarios, for example, the target objects in the tracking area have different sizes, there are small-sized target objects in the tracking area that are difficult to detect, the target objects in the tracking area are dense, and the target objects are too small to be detected. Target tracking scenarios such as similar appearance, high motion complexity of the target object, and severe occlusion.
在一种可能的实现方式中，在步骤S11之前，所述方法还包括：获取标志物的第一坐标；对于所述多个摄像头中的任一摄像头，根据该摄像头获得的所述标志物的第二坐标，以及所述标志物的第一坐标，确定该摄像头对应的转换矩阵。In a possible implementation manner, before step S11, the method further includes: acquiring the first coordinates of the marker; for any camera in the plurality of cameras, obtaining the marker's coordinates according to the camera The second coordinate, and the first coordinate of the marker, determine the transformation matrix corresponding to the camera.
在该实现方式中，在对所述多个摄像头进行标定时，可以先在跟踪区域中以一定的密度放置标志物。图5示出在本公开实施例中在跟踪区域中放置标志物的示意图。在图5所示的示例中，跟踪区域是足球场，标志物是白色的金属片。当然，本领域技术人员可以根据实际应用场景需求和/或个人喜好灵活选择标志物，这里，按照一定密度放置的标志物可以是不同的高度、不同的大小、不同的颜色等在此不作限定。通过采用标志物进行标定，能够提高在角点不丰富的跟踪区域中进行标定的准确性，从而有助于提高目标跟踪的准确性。In this implementation, when calibrating the plurality of cameras, markers may be placed in the tracking area at a certain density first. FIG. 5 shows a schematic diagram of placing markers in a tracking area in an embodiment of the present disclosure. In the example shown in Figure 5, the tracking area is a football field and the marker is a white metal sheet. Of course, those skilled in the art can flexibly select markers according to actual application scenario requirements and/or personal preferences. Here, markers placed according to a certain density can be of different heights, different sizes, and different colors, which are not limited here. By using markers for calibration, the accuracy of the calibration in the tracking area without rich corners can be improved, thereby helping to improve the accuracy of target tracking.
作为该实现方式的一个示例，第一坐标可以是世界坐标系下的坐标，第二坐标可以是像素坐标系下的坐标，转换矩阵可以是单应矩阵。由于目标对象在同一平面内运动，因此，可以采用公式(1)确定转换矩阵H 3×3： As an example of this implementation, the first coordinates may be coordinates in a world coordinate system, the second coordinates may be coordinates in a pixel coordinate system, and the transformation matrix may be a homography matrix. Since the target object moves in the same plane, formula (1) can be used to determine the transformation matrix H 3×3 :
Figure PCTCN2021086020-appb-000001
Figure PCTCN2021086020-appb-000001
其中，(X w,Y ω)表示标志物的第一坐标，例如可以是标志物在世界坐标系下的坐标；(u,v)表示摄像 头获得的标志物的第二坐标，例如可以是标志物在像素坐标系下的坐标。 Among them, (X w , Y ω ) represents the first coordinate of the marker, for example, it can be the coordinate of the marker in the world coordinate system; (u, v) represents the second coordinate of the marker obtained by the camera, for example, it can be the marker The coordinates of the object in the pixel coordinate system.
在该示例中，单应矩阵H 3×3的自由度是8，理论上采用4个标志物对应的特征点就可以求解出所述单应矩阵。为了提高内外参的精度，如图5所示，可以设置更多的标志物，采用最小二乘法，并过滤孤立点，利用棋盘格标定的方法得到单应矩阵。由此得到的单应矩阵能够去除摄像头采集的图像中的畸变，并能提高所确定的第一坐标的准确性。 In this example, the degree of freedom of the homography matrix H 3×3 is 8, and theoretically, the homography matrix can be solved by using the feature points corresponding to 4 markers. In order to improve the accuracy of internal and external parameters, as shown in Figure 5, more markers can be set, the least squares method can be used, outliers can be filtered, and the homography matrix can be obtained by using the checkerboard calibration method. The homography matrix thus obtained can remove the distortion in the image captured by the camera, and can improve the accuracy of the determined first coordinate.
这里，可以通过直接线性转换(Direct Linear Transformation，DLT)和奇异值分解(Singular Value Decomposition，SVD)获取摄像头外参，通过棋盘格标定获取摄像头内参，进而去除图像畸变。Here, the camera external parameters can be obtained through Direct Linear Transformation (DLT) and Singular Value Decomposition (SVD), and the camera internal parameters can be obtained through checkerboard calibration to remove image distortion.
在该实现方式中，通过确定所述多个摄像头中的各个摄像头对应的转换矩阵，由此能够将所述多个摄像头获得的目标对象的第二坐标转换到统一的坐标系中，例如，转换到世界坐标系中，得到目标对象对应于所述多个摄像头中的各个摄像头的第一坐标，由此能够便于后续对所述目标对象进行跟踪。In this implementation manner, by determining the transformation matrix corresponding to each of the multiple cameras, the second coordinates of the target object obtained by the multiple cameras can be transformed into a unified coordinate system, for example, by converting In the world coordinate system, the first coordinates of the target object corresponding to each of the plurality of cameras are obtained, thereby facilitating subsequent tracking of the target object.
在一种可能的实现方式中，所述方法还包括：对于任一目标对象，确定该目标对象对应于多个摄像头的第一坐标；根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，其中，所述跟踪区域包括多个子区域；根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头的第一坐标进行融合，得到该目标对象的融合坐标；在所述根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象之前，还包括：根据所述跟踪区域中目标对象的融合坐标，确定所述跟踪区域中目标对象之间的距离。In a possible implementation manner, the method further includes: for any target object, determining that the target object corresponds to the first coordinates of the plurality of cameras; according to the first coordinates of the target object corresponding to the plurality of cameras , determine the sub-area where the target object is located in the tracking area, wherein the tracking area includes multiple sub-areas; The object is fused corresponding to the first coordinates of the plurality of cameras, and the fusion coordinates of the target object are obtained; before the plurality of conflicting target objects are determined according to the distance between the target objects in the tracking area, the method further includes: The distance between the target objects in the tracking area is determined according to the fusion coordinates of the target objects in the tracking area.
在该实现方式中，任一目标对象对应于任一摄像头的第一坐标，可以表示根据该摄像头采集的图像得到的该目标对象的第一坐标。第一坐标可以是第一坐标系下的坐标，例如，第一坐标系可以是世界坐标系或者其他虚拟坐标系。融合坐标和第一坐标可以是相同坐标系下的坐标，例如，融合坐标和第一坐标可以都是世界坐标系下的坐标。In this implementation manner, any target object corresponds to the first coordinates of any camera, and may represent the first coordinates of the target object obtained according to the images collected by the camera. The first coordinates may be coordinates in a first coordinate system, for example, the first coordinate system may be a world coordinate system or other virtual coordinate systems. The fusion coordinates and the first coordinates may be coordinates in the same coordinate system, for example, the fusion coordinates and the first coordinates may both be coordinates in the world coordinate system.
作为该实现方式的一个示例，预先确定所述跟踪区域中的各个子区域在第一坐标系下的坐标范围。例如，子区域为矩形，任一子区域在第一坐标系下的坐标范围可以采用该子区域在第一坐标系下的四个顶点的坐标来表示，或者，该子区域在第一坐标系下的坐标范围可以采用该子区域在第一坐标系下的左上角顶点的坐标以及该子区域的宽度和高度来表示。当然，任一子区域在第一坐标系下的坐标范围还可以采用其他方式来表示，在此不作限定。另外，任一子区域的形状也可以不为矩形，例如可以为三角形等。跟踪区域中的不同子区域的大小可以相同，也可以不同。As an example of this implementation, the coordinate range of each sub-area in the tracking area in the first coordinate system is predetermined. For example, if the sub-region is a rectangle, the coordinate range of any sub-region in the first coordinate system can be represented by the coordinates of the four vertices of the sub-region in the first coordinate system, or, the sub-region in the first coordinate system The coordinate range below can be represented by the coordinates of the top-left corner vertex of the sub-region in the first coordinate system and the width and height of the sub-region. Certainly, the coordinate range of any sub-region under the first coordinate system may also be represented by other manners, which is not limited here. In addition, the shape of any sub-region may not be a rectangle, for example, a triangle or the like. Different sub-areas in the tracking area can be the same size or different.
在该实现方式中，根据目标对象对应于所述多个摄像头的第一坐标，以及子区域在第一坐标系下的坐标范围，可以确定该目标对象在所述跟踪区域中所处的子区域。在该实现方式中，可以根据目标对象对应于所述多个摄像头中的全部或部分摄像头的第一坐标，以及子区域在第一坐标系下的坐标范围，确定该目标对象在所述跟踪区域中所处的子区域。In this implementation manner, according to the first coordinates of the target object corresponding to the plurality of cameras, and the coordinate range of the sub-region in the first coordinate system, the sub-region where the target object is located in the tracking region can be determined . In this implementation manner, it may be determined that the target object is in the tracking area according to the first coordinates of the target object corresponding to all or part of the cameras in the plurality of cameras, and the coordinate range of the sub-region in the first coordinate system the sub-region in which it is located.
作为该实现方式的一个示例，根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，包括：根据该目标对象对应于所述多个摄像头中的任一摄像头的第一坐标，以及子区域在第一坐标系下的坐标范围，确定该目标对象在所述跟踪区域中所处的候选子区域；将得票数最高的候选子区域确定为该目标对象所处的子区域。例如，所述多个摄像头包括摄像头C 1、摄像头C 2、摄像头C 3和摄像头C 4；根据该目标对象对应于摄像头C 1的第一坐标，确定该目标对象在所述跟踪区域中所处的候选子区域是子区域A 1；根据该目标对象对应于摄像头C 2的第一坐标，确定该目标对象在所述跟踪区域中所处的候选子区域是子区域A 1；根据该目标对象对应于摄像头C 3的第一坐标，确定该目标对象在所述跟踪区域中所处的候选子区域是子区域A 2；根据该目标对象对应于摄像头C 4的第一坐标，确定该目标对象在所述跟踪区域中所处的候选子区域是子区域A 1；那么，候选子区域A 1的得票数是3，候选子区域A 2的得票数是1，因此，可以将候选子区域A 1确定为该目标对象所处的子区域。 As an example of this implementation, determining the sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras includes: according to the target object corresponding to the The first coordinates of any camera in the plurality of cameras, and the coordinate range of the sub-region in the first coordinate system, determine the candidate sub-region where the target object is located in the tracking region; the candidate sub-region with the highest number of votes is determined. The area is determined as the sub-area where the target object is located. For example, the plurality of cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4 ; according to the first coordinate of the target object corresponding to the camera C 1 , determine where the target object is located in the tracking area The candidate sub-region is sub-region A 1 ; according to the first coordinate of the target object corresponding to the camera C 2 , it is determined that the candidate sub-region where the target object is located in the tracking region is sub-region A 1 ; according to the target object Corresponding to the first coordinate of the camera C3, it is determined that the candidate sub-region where the target object is located in the tracking area is the sub-region A2 ; according to the first coordinate of the target object corresponding to the camera C4 , determine the target object The candidate sub-area located in the tracking area is the sub-area A 1 ; then, the number of votes for the candidate sub-area A 1 is 3, and the number of votes for the candidate sub-area A 2 is 1. Therefore, the candidate sub-area A can be 1 is determined as the sub-area where the target object is located.
作为该实现方式的另一个示例，根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，包括：根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象对应于所述多个摄像头的第一坐标的平均值；根据子区域在第一坐标系下的坐标范围，确定所述第一坐标的平均值所处的子区域，并将所述第一坐标的平均值所处的子区域确定为该目标对象在所述跟踪区域中所处的子区域。As another example of the implementation manner, determining the sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras includes: according to the target object corresponding to the The first coordinates of the multiple cameras are determined, and the average value of the first coordinates of the target object corresponding to the multiple cameras is determined; according to the coordinate range of the sub-region in the first coordinate system, the average value of the first coordinates is determined The sub-area where the target object is located, and the sub-area where the average value of the first coordinates is located is determined as the sub-area where the target object is located in the tracking area.
在该实现方式中，可以根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头中的全部或部分摄像头的第一坐标进行融合，得到该目标对象的融合坐标。例如，所述多个摄像头包括摄像头C 1、摄像头C 2、摄像头C 3和摄像头C 4，该目标对象处于子区域A 1，摄像头C 1针对子区域A 1的置信度是Z 11，摄像头C 2针对子区域A 1的置信度是Z 21，摄像头C 3针对子区域A 1的置信度是Z 31，摄像头C 4针对子区域A 1的置信度是Z 41，该目标对象对应于摄像头C 1的第一坐标是P w1，该目标对象对应于摄像头C 2的第一坐标是P w2，该目标对象对应于摄像头C 3的第一坐标是P w3，该目标对象对应于摄像头C 4的第一坐标是P w4，则该目标对象的融合坐标可以为
Figure PCTCN2021086020-appb-000002
又如，在所述多个摄像头中，可以选取针对该目标对象所处的子区域的置信度较大的部分摄像头，确定该目标对象的融合坐标。例如，Z 11＞Z 21＞Z 31＞Z 41，
Figure PCTCN2021086020-appb-000003
In this implementation manner, the first coordinates of the target object corresponding to all or part of the cameras in the plurality of cameras may be fused according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located, Get the fusion coordinates of the target object. For example, the multiple cameras include camera C 1 , camera C 2 , camera C 3 and camera C 4 , the target object is located in sub-region A 1 , the confidence level of camera C 1 for sub-region A 1 is Z 11 , and camera C 2 The confidence level for sub-region A 1 is Z 21 , the confidence level for camera C 3 for sub-region A 1 is Z 31 , the confidence level for camera C 4 for sub-region A 1 is Z 41 , the target object corresponds to camera C The first coordinate of 1 is P w1 , the first coordinate of the target object corresponding to the camera C 2 is P w2 , the first coordinate of the target object corresponding to the camera C 3 is P w3 , the target object corresponding to the camera C 4 The first coordinate is P w4 , then the fusion coordinate of the target object can be
Figure PCTCN2021086020-appb-000002
For another example, among the plurality of cameras, a part of the cameras with higher confidence in the sub-region where the target object is located may be selected to determine the fusion coordinates of the target object. For example, Z 11 > Z 21 > Z 31 > Z 41 ,
Figure PCTCN2021086020-appb-000003
在该实现方式中，通过结合目标对象对应于所述多个摄像头的第一坐标，得到该目标对象的融合坐标，并基于该目标对象的融合坐标进行目标跟踪，由此能够得到基于更精准的坐标进行目标跟踪，从而有助于提高目标跟踪的准确性。In this implementation, by combining the first coordinates of the target object corresponding to the plurality of cameras, the fusion coordinates of the target object are obtained, and the target tracking is performed based on the fusion coordinates of the target object, so that a more accurate The coordinates are used for target tracking, which helps to improve the accuracy of target tracking.
在一种可能的实现方式中，对于任一目标对象和所述多个摄像头中的任一摄像头，根据该摄像头获得的该目标对象的第二坐标，以及该摄像头对应的转换矩阵，确定该目标对象对应于该摄像头的第一坐标。In a possible implementation manner, for any target object and any camera among the plurality of cameras, the target is determined according to the second coordinates of the target object obtained by the camera and the transformation matrix corresponding to the camera The object corresponds to the camera's first coordinates.
作为该实现方式的一个示例，该摄像头获得的该目标对象的第二坐标，可以是该目标对象在该摄像头对应的像素坐标系下的坐标。即，第二坐标可以表示像素坐标系下的坐标。在该示例中，该目标对象对应于该摄像头的第一坐标，表示根据该摄像头获得的该目标对象第二坐标确定的该目标对象在世界坐标系下的坐标。在该示例中，该摄像头对应的转换矩阵可以是单应矩阵。As an example of this implementation, the second coordinate of the target object obtained by the camera may be the coordinates of the target object in a pixel coordinate system corresponding to the camera. That is, the second coordinates may represent coordinates in the pixel coordinate system. In this example, the target object corresponds to the first coordinates of the camera, and represents the coordinates of the target object in the world coordinate system determined according to the second coordinates of the target object obtained by the camera. In this example, the transformation matrix corresponding to the camera may be a homography matrix.
在其他示例中，该摄像头获得的该目标对象的第二坐标，还可以是该目标对象在该摄像头对应的图像坐标系等坐标系下的坐标。即，第二坐标还可以是图像坐标系等坐标系下的坐标。In other examples, the second coordinate of the target object obtained by the camera may also be the coordinates of the target object in a coordinate system such as an image coordinate system corresponding to the camera. That is, the second coordinates may also be coordinates in a coordinate system such as an image coordinate system.
作为该实现方式的一个示例，可以对该摄像头采集的图像进行目标检测，确定目标对象在所述图像中的检测框，并根据检测框的位置，确定该摄像头获得的该目标对象的第二坐标。在该示例中，可以将检测框上的任意一点或者检测框内部的任意一点作为该目标对象的第二坐标。例如，可以将检测框底边的中点作为该目标对象的第二坐标。As an example of this implementation, target detection may be performed on an image collected by the camera, a detection frame of the target object in the image may be determined, and according to the position of the detection frame, the second coordinates of the target object obtained by the camera may be determined . In this example, any point on the detection frame or any point inside the detection frame may be used as the second coordinate of the target object. For example, the midpoint of the bottom edge of the detection frame can be used as the second coordinate of the target object.
例如，所述多个摄像头中的任一摄像头对应的转换矩阵为H，该摄像头获得的任一目标对象的第二坐标为P u，则该目标对象对应于该摄像头的第一坐标可以采用公式(2)确定： For example, the transformation matrix corresponding to any camera in the plurality of cameras is H, and the second coordinate of any target object obtained by the camera is P u , then the first coordinate of the target object corresponding to the camera can use the formula (2) Determine:
P w＝(H TH) -1H TP u               (2) P w = (H T H) -1 H T P u (2)
在一种可能的实现方式中，所述方法还包括：对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度。In a possible implementation manner, the method further includes: for any camera in the plurality of cameras, according to the distance between the camera and the plurality of sub-regions of the tracking area, determine that the camera is aimed at the tracking area. Confidence for multiple subregions.
在该实现方式中，可以根据所述多个摄像头与所述跟踪区域的多个子区域之间的距离，确定所述多个摄像头针对所述多个子区域的置信度。在该实现方式中，任一摄像头针对任一子区域的置信度，与该摄像头与该子区域之间的距离负相关。即，该摄像头与该子区域之间的距离越大，则该摄像头针对该子区域的置信度越低；该摄像头与该子区域之间的距离越小，则该摄像头针对该子区域的置信度越高。例如，若摄像头C 1与子区域A 1之间的距离，小于摄像头C 2与子区域A 1之间的距离，则摄像头C 1针对子区域A 1的置信度Z 11，可以大于摄像头C 2针对子区域A 1的置信度Z 21。又如，若摄像头C 1与子区域A 1之间的距离，小于摄像头C 1与子区域A 2之间的距离，则摄像头C 1针对子区域A 1的置信度Z 11，可以大于摄像头C 1针对子区域A 2的置信度Z 12。 In this implementation manner, the confidence levels of the multiple cameras with respect to the multiple sub-areas may be determined according to the distances between the multiple cameras and the multiple sub-areas of the tracking area. In this implementation, the confidence of any camera with respect to any sub-region is negatively correlated with the distance between the camera and the sub-region. That is, the larger the distance between the camera and the sub-area, the lower the confidence of the camera in the sub-area; the smaller the distance between the camera and the sub-area, the lower the confidence of the camera in the sub-area higher degree. For example, if the distance between the camera C 1 and the sub-area A 1 is smaller than the distance between the camera C 2 and the sub-area A 1 , the confidence Z 11 of the camera C 1 for the sub-area A 1 may be greater than that of the camera C 2 Confidence Z 21 for sub-region A 1 . For another example, if the distance between the camera C 1 and the sub-area A 1 is smaller than the distance between the camera C 1 and the sub-area A 2 , then the confidence Z 11 of the camera C 1 for the sub-area A 1 may be greater than that of the camera C 1 Confidence Z 12 for sub-region A 2 .
在该实现方式中，对于所述多个摄像头中的任一摄像头，该摄像头与任一子区域之间的距离越小，则该摄像头拍摄的该子区域的视频帧越清晰，从而越有利于准确识别视频帧中的信息。相反，该摄像头与任一子区域之间的距离越大，则该摄像头拍摄的该子区域的视频帧越模糊，从而越不利于准确识别视频帧中的信息。因此，在该实现方式中，对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度，由此将摄像头与子区域之间的距离作为摄像头针对子区域的置信度的考虑因素。基于由此确定的置信度进行坐标融合和/或确定用于提取目标对象的当前特征的摄像头，从而有助于提高目标跟踪的准确性。In this implementation manner, for any camera among the plurality of cameras, the smaller the distance between the camera and any sub-area, the clearer the video frame of the sub-area captured by the camera, which is more beneficial to Accurately identify information in video frames. On the contrary, the larger the distance between the camera and any sub-area, the more blurred the video frame of the sub-area captured by the camera, which is not conducive to accurately identifying the information in the video frame. Therefore, in this implementation manner, for any camera among the plurality of cameras, the confidence level of the camera for the plurality of sub-areas is determined according to the distance between the camera and the plurality of sub-areas of the tracking area, The distance between the camera and the sub-area is thus taken into consideration for the confidence of the camera for the sub-area. Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
在一种可能的实现方式中，在所述确定所述多个摄像头针对所述多个子区域的置信度之后，所述方法还包括：对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，调整该摄像头针对所述多个子区域的置信度。In a possible implementation manner, after the determining the confidence of the plurality of cameras with respect to the plurality of sub-regions, the method further includes: for any camera in the plurality of cameras, according to the camera The average distance between the target objects in the collected video frames is used to adjust the confidence of the camera for the multiple sub-regions.
作为该实现方式的一个示例，可以在目标跟踪的过程中，根据该摄像头采集的视频帧中目标对象之间的平均距离，不断调整该摄像头针对所述多个子区域的置信度。例如，可以以预设频率调整该摄像头针对所述多个子区域的置信度。例如，预设频率可以是1秒。As an example of this implementation, during the target tracking process, the confidence of the camera for the multiple sub-regions may be continuously adjusted according to the average distance between the target objects in the video frames collected by the camera. For example, the confidence of the camera with respect to the plurality of sub-regions may be adjusted at a preset frequency. For example, the preset frequency may be 1 second.
在本公开实施例中，可以对多个摄像头针对该目标对象所处的子区域的置信度进行加权处理，再根据加权处理后的多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于多个摄像头的第一坐标进行融合。In the embodiment of the present disclosure, a weighting process may be performed on the confidences of multiple cameras with respect to the sub-region where the target object is located, and then according to the weighted confidences of the multiple cameras with respect to the sub-region where the target object is located, The first coordinates of the target object corresponding to the multiple cameras are fused.
对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中所有目标对象两两之间的平均 距离，调整该摄像头针对所述多个子区域的置信度。例如，所述多个摄像头包括摄像头C 1、摄像头C 2、摄像头C 3和摄像头C 4，摄像头C 1采集的视频帧中目标对象之间的平均距离是D 1，摄像头C 2采集的视频帧中目标对象之间的平均距离是D 2，摄像头C 3采集的视频帧中目标对象之间的平均距离是D 3，摄像头C 4采集的视频帧中目标对象之间的平均距离是D 4，D 1＞D 2＞D 3＞D 4，则可以将摄像头C 1针对所述多个子区域中的各个子区域的置信度增加B 1，将摄像头C 2针对所述多个子区域中的各个子区域的置信度增加B 2，将摄像头C 3针对所述多个子区域中的各个子区域的置信度增加B 3，将摄像头C 4针对所述多个子区域中的各个子区域的置信度保持不变，其中，B 1＞B 2＞B 3＞0，例如，B 1＝1.5，B 2＝1，B 3＝0.5。 For any camera in the plurality of cameras, according to the average distance between all target objects in the video frames collected by the camera, the confidence level of the camera with respect to the plurality of sub-regions is adjusted. For example, the multiple cameras include a camera C 1 , a camera C 2 , a camera C 3 and a camera C 4 , the average distance between target objects in the video frames collected by the camera C 1 is D 1 , and the video frames collected by the camera C 2 The average distance between the target objects is D 2 , the average distance between the target objects in the video frame collected by the camera C 3 is D 3 , and the average distance between the target objects in the video frame collected by the camera C 4 is D 4 , D 1 >D 2 >D 3 >D 4 , then the confidence of the camera C 1 for each of the sub-regions in the plurality of sub-regions can be increased by B 1 , and the camera C 2 for each of the sub-regions in the plurality of sub-regions. The confidence of the region is increased by B 2 , the confidence of the camera C 3 for each of the sub-regions is increased by B 3 , and the confidence of the camera C 4 for each of the sub-regions is kept unchanged. variable, where B 1 >B 2 >B 3 >0, eg, B 1 =1.5, B 2 =1, and B 3 =0.5.
在该实现方式中，对于所述多个摄像头中的任一摄像头，该摄像头采集的视频帧中目标对象之间的平均距离越小，则该摄像头采集的视频帧中的目标对象之间相互遮挡的可能性越大，从而不利于对目标对象进行特征提取，进而不利于进行目标跟踪。相反，该摄像头采集的视频帧中目标对象之间的平均距离越大，则该摄像头采集的视频帧中的目标对象之间相互遮挡的可能性越小，从而有利于对目标对象进行特征提取，进而有利于进行目标跟踪。在该实现方式中，通过对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，动态调整该摄像头针对所述多个子区域的置信度，并基于由此确定的置信度进行坐标融合和/或确定用于提取目标对象的当前特征的摄像头，从而有助于提高目标跟踪的准确性。In this implementation, for any camera in the plurality of cameras, the smaller the average distance between the target objects in the video frame collected by the camera, the smaller the average distance between the target objects in the video frame collected by the camera. The greater the possibility is, it is not conducive to the feature extraction of the target object, which is not conducive to the target tracking. On the contrary, the larger the average distance between the target objects in the video frames collected by the camera, the less likely the target objects in the video frames collected by the camera are occluded from each other, which is conducive to the feature extraction of the target objects. This facilitates target tracking. In this implementation, for any camera in the plurality of cameras, according to the average distance between the target objects in the video frames collected by the camera, dynamically adjust the confidence level of the camera with respect to the plurality of sub-regions, and Coordinate fusion and/or determination of a camera for extracting current features of the target object is performed based on the confidence level thus determined, thereby helping to improve the accuracy of target tracking.
在一种可能的实现方式中，所述方法还包括：对于任一目标对象，响应于该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突的情况，将所述上一时刻不与其他目标对象冲突、且未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；或者，响应于该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况，将该目标对象在所述上一时刻所属的冲突区中、未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息。In a possible implementation manner, the method further includes: for any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment The target object that does not conflict with other target objects at the last moment and does not match other target objects at the current moment is determined as the remaining target objects at the last moment; or, in response to the target object being in the current moment When the time and the previous time conflict with the same other target object, the target object that does not match with other target objects at the current time in the conflict zone to which the target object belongs to the previous time is determined as The remaining target objects at the last moment; from the remaining target objects at the last moment, the target object with the closest distance to the target object at the current moment is determined to match the target object at the current moment The target object, and the identification information of the matching target object is used as the identification information of the target object.
例如，某一目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突，且所述上一时刻不与其他目标对象冲突的目标对象有15个，分别是目标对象O 1至O 15。若上一时刻中的目标对象O 1和O 2已与当前时刻的其他目标对象匹配，即，标识信息O 1和O 2已作为当前时刻的目标对象的标识信息，则可以将目标对象O 3至O 15确定为上一时刻的剩余目标对象。在确定上一时刻的剩余目标对象O 3至O 15之后，从上一时刻的剩余目标对象O 3至O 15中，将与当前时刻的该目标对象距离最近的目标对象，确定为与当前时刻的该目标对象匹配的目标对象。例如，若在上一时刻的剩余目标对象O 3至O 15中，O 6与当前时刻的该目标对象距离最近，则将上一时刻的目标对象O 6确定为与当前时刻的该目标对象匹配的目标对象，并将当前时刻的该目标对象的标识信息确定为O 6。 For example, a certain target object does not conflict with other target objects at the current moment and the last moment of the current moment, and there are 15 target objects that do not conflict with other target objects at the last moment, which are target objects 0 1 to O 15 . If the target objects O 1 and O 2 in the previous moment have been matched with other target objects at the current moment, that is, the identification information O 1 and O 2 have been used as the identification information of the target objects at the current moment, the target object O 3 To O 15 is determined as the remaining target object at the previous moment. After the remaining target objects O 3 to O 15 at the previous moment are determined, from the remaining target objects O 3 to O 15 at the previous moment, the target object with the closest distance to the target object at the current moment is determined as the target object at the current moment. The target object that matches this target object. For example, if among the remaining target objects O 3 to O 15 at the previous moment, O 6 is the closest to the target object at the current moment, then the target object O 6 at the previous moment is determined to match the target object at the current moment. the target object, and the identification information of the target object at the current moment is determined as O 6 .
又如，某一目标对象在当前时刻和所述上一时刻与相同的其他目标对象冲突，例如，该目标对象在所述上一时刻所属的冲突区包括目标对象O 1、目标对象O 2、目标对象O 3和目标对象O 4。若在该目标对象在所述上一时刻所属的冲突区中，目标对象O 1和目标对象O 2已与当前时刻的其他目标对象匹配，即，标识信息O 1和O 2已作为当前时刻的目标对象的标识信息，则可以将目标对象O 3和目标对象O 4确定为上一时刻的剩余目标对象。在确定上一时刻的剩余目标对象O 3和O 4之后，从上一时刻的剩余目标对象O 3和O 4中，将与当前时刻的该目标对象距离最近的目标对象，确定为与当前时刻的该目标对象匹配的目标对象。例如，若在上一时刻的剩余目标对象O 3和O 4中，O 4与当前时刻的该目标对象距离最近，则将上一时刻的目标对象O 4确定为与当前时刻的该目标对象匹配的目标对象，并将当前时刻的该目标对象的标识信息确定为O 4。 For another example, a certain target object conflicts with the same other target objects at the current moment and the last moment. For example, the conflict area to which the target object belongs at the last moment includes target object O 1 , target object O 2 , The target object O 3 and the target object O 4 . If in the conflict zone to which the target object belongs at the last moment, the target object O1 and the target object O2 have been matched with other target objects at the current moment, that is, the identification information O1 and O2 have been used as the current moment's The identification information of the target object, then the target object O 3 and the target object O 4 can be determined as the remaining target objects at the previous moment. After determining the remaining target objects O 3 and O 4 at the previous moment, from the remaining target objects O 3 and O 4 at the previous moment, the target object with the closest distance to the target object at the current moment is determined as the target object with the current moment. The target object that matches this target object. For example, if among the remaining target objects O 3 and O 4 at the previous moment, O 4 is the closest to the target object at the current moment, then the target object O 4 at the previous moment is determined to match the target object at the current moment. the target object, and the identification information of the target object at the current moment is determined as O 4 .
在该实现方式中，对于任一目标对象，在该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突或者该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况下，从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息，由此采用基于距离的贪心算法来确定该目标对象的标识信息，而不考虑该目标对象的视觉特征，从而能够减少计算量，降低时间开销，满足目标跟踪的实时性要求。In this implementation manner, for any target object, the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, or the target object is at the current moment and the previous moment. In the case of conflict with the same other target object, from the remaining target objects at the previous moment, the target object with the closest distance to the target object at the current moment is determined as the target object at the current moment. The matching target object, and the identification information of the matching target object is used as the identification information of the target object, thereby adopting the distance-based greedy algorithm to determine the identification information of the target object, regardless of the visual feature of the target object, Thereby, the amount of calculation can be reduced, the time overhead can be reduced, and the real-time requirements of target tracking can be met.
在一种可能的实现方式中，本公开实施例提供的目标跟踪方法可以采用目标跟踪模型来实现。例如，所述目标跟踪模型可以采用Faster-RCNN(Faster Recurrent Convolutional Neural Network，更快速的递归卷积神经网络)或者Fast-RCNN(Faster Recurrent Convolutional Neural Network，快速递归卷积神经网络)等。所述目标跟踪模型的骨干网络可以采用ResNet-18等结构。为了提高所述目标跟踪模型的处理速度，可以通过通道剪枝的方法对所述目标跟踪模型进行模型压缩，同时，采用多层特征检测的方法提 高模型精度，由此在不降低模型精度的同时，提高模型速度。In a possible implementation manner, the target tracking method provided by the embodiment of the present disclosure may be implemented by using a target tracking model. For example, the target tracking model may adopt Faster-RCNN (Faster Recurrent Convolutional Neural Network, faster recurrent convolutional neural network) or Fast-RCNN (Faster Recurrent Convolutional Neural Network, fast recursive convolutional neural network) or the like. The backbone network of the target tracking model can adopt structures such as ResNet-18. In order to improve the processing speed of the target tracking model, the target tracking model can be compressed by channel pruning, and at the same time, the multi-layer feature detection method is used to improve the model accuracy, so that the model accuracy is not reduced at the same time. , to increase model speed.
在一种可能的实现方式中，所述方法还可以包括：输出目标对象的标识信息与坐标之间的对应关系。在一个例子中，可以按照标识信息的顺序，输出标识信息与坐标之间的对应关系。例如，标识信息包括球衣号码1-11，则可以按照球衣号码由小到大的顺序，输出对应的目标对象的坐标。In a possible implementation manner, the method may further include: outputting the correspondence between the identification information of the target object and the coordinates. In one example, the correspondence between the identification information and the coordinates may be output in the order of the identification information. For example, if the identification information includes the jersey numbers 1-11, the coordinates of the corresponding target objects can be output according to the order of the jersey numbers from small to large.
在一种可能的实现方式中，可以根据目标对象在多个时刻(例如多个连续时刻)的坐标，得到目标对象的运动轨迹。在一个示例中，还可以采用卡尔曼滤波算法对目标对象的运动轨迹进行处理，以使目标对象的运动轨迹更平滑。In a possible implementation manner, the motion trajectory of the target object may be obtained according to the coordinates of the target object at multiple times (for example, multiple consecutive times). In an example, a Kalman filter algorithm may also be used to process the motion trajectory of the target object, so as to make the motion trajectory of the target object smoother.
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。It can be understood that the above-mentioned method embodiments mentioned in the present disclosure can be combined with each other to form a combined embodiment without violating the principle and logic.
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。Those skilled in the art can understand that in the above method of the specific implementation, the writing order of each step does not mean a strict execution order but constitutes any limitation on the implementation process, and the specific execution order of each step should be based on its function and possible Internal logic is determined.
此外，本公开还提供了目标跟踪装置、电子设备、计算机可读存储介质、程序，上述均可用来实现本公开提供的任一种目标跟踪方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。In addition, the present disclosure also provides target tracking devices, electronic devices, computer-readable storage media, and programs, all of which can be used to implement any target tracking method provided by the present disclosure. For the corresponding technical solutions and descriptions, refer to the corresponding records in the Methods section. ,No longer.
图6示出本公开实施例提供的目标跟踪装置的框图。如图6所示，所述目标跟踪装置包括：FIG. 6 shows a block diagram of a target tracking apparatus provided by an embodiment of the present disclosure. As shown in Figure 6, the target tracking device includes:
第一确定部分61，被配置为根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象；The first determining part 61 is configured to determine a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area;
提取部分62，被配置为响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征；The extraction part 62 is configured to extract the current feature of the target object in response to the situation that any target object in the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects;
第二确定部分63，被配置为基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目标对象提取的特征；The second determining part 63 is configured to determine, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object, wherein the plurality of reference features are in the plurality of target objects. Before the target objects collide with each other, the features extracted from the multiple target objects respectively;
第三确定部分64，被配置为将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。The third determining part 64 is configured to determine, as the identification information of the target object, the identification information corresponding to the reference feature matching the current feature of the target object.
在一种可能的实现方式中，彼此冲突的多个目标对象中的任一目标对象，与所述多个目标对象中的其他至少一个目标对象之间的距离小于或等于距离阈值；In a possible implementation manner, the distance between any target object among the multiple target objects that conflict with each other and at least one other target object among the multiple target objects is less than or equal to a distance threshold;
所述提取部分62被配置为：The extraction section 62 is configured to:
对于所述多个目标对象中的任一目标对象，响应于该目标对象与所述多个目标对象中的其他目标对象之间的距离均大于所述距离阈值的情况，提取该目标对象的当前特征。For any target object in the plurality of target objects, in response to the situation that the distance between the target object and other target objects in the plurality of target objects is greater than the distance threshold, extract the current value of the target object feature.
在一种可能的实现方式中，所述装置还包括：In a possible implementation, the apparatus further includes:
第三确定部分64，被配置为对于任一目标对象，确定该目标对象对应于多个摄像头的第一坐标；其中，该目标对象对应于任一摄像头的第一坐标用于表征根据该摄像头采集的图像得到的该目标对象的第一坐标；The third determination part 64 is configured to, for any target object, determine the first coordinates of the target object corresponding to the plurality of cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the acquisition according to the camera. The first coordinate of the target object obtained from the image of ;
第四确定部分，被配置为根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，其中，所述跟踪区域包括多个子区域；a fourth determining part, configured to determine a sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras, wherein the tracking area includes a plurality of sub-areas ;
融合部分，被配置为根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头的第一坐标进行融合，得到该目标对象的融合坐标；The fusion part is configured to fuse the first coordinates of the target object corresponding to the plurality of cameras according to the confidence of the plurality of cameras for the sub-region where the target object is located to obtain the fusion coordinates of the target object ;
第五确定部分，被配置为根据所述跟踪区域中目标对象的融合坐标，确定所述跟踪区域中目标对象之间的距离。The fifth determination part is configured to determine the distance between the target objects in the tracking area according to the fusion coordinates of the target objects in the tracking area.
在一种可能的实现方式中，所述提取部分62被配置为：In a possible implementation, the extraction part 62 is configured as:
确定该目标对象在所述跟踪区域中所处的子区域；determining the sub-area where the target object is located in the tracking area;
根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头；Determine the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located;
根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征。Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
在一种可能的实现方式中，所述提取部分62被配置为：In a possible implementation, the extraction part 62 is configured as:
根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头。According to the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the The camera that extracts the current features of this target object.
在一种可能的实现方式中，所述重叠信息包括该目标对象的检测框与其他目标对象的检测框的交并比；In a possible implementation manner, the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
所述提取部分62被配置为：The extraction section 62 is configured to:
从满足重叠条件的摄像头中，将针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，其中，所述重叠条件表示在所述当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值。From the cameras that satisfy the overlapping condition, the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition The intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
在一种可能的实现方式中，所述装置还包括：In a possible implementation, the apparatus further includes:
第六确定部分，被配置为对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度。The sixth determining part is configured to, for any one of the plurality of cameras, determine the confidence of the camera for the plurality of sub-areas according to the distance between the camera and the plurality of sub-areas of the tracking area.
在一种可能的实现方式中，所述装置还包括：In a possible implementation, the apparatus further includes:
调整部分，被配置为对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，调整该摄像头针对所述多个子区域的置信度。The adjusting part is configured to, for any one of the plurality of cameras, adjust the confidence of the camera for the plurality of sub-regions according to the average distance between the target objects in the video frames collected by the camera.
在一种可能的实现方式中，所述第二确定部分63被配置为：In a possible implementation manner, the second determining part 63 is configured as:
从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征；From the plurality of reference features corresponding to the plurality of target objects, determine the reference feature with the greatest similarity to the current feature of the target object;
响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。In response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold, determine the reference feature with the greatest similarity as the reference matching the current feature of the target object feature.
在一种可能的实现方式中，所述装置还包括：In a possible implementation, the apparatus further includes:
第七确定部分，被配置为响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征。A seventh determination part, configured to respond to a situation where the similarity between the reference feature with the greatest similarity and the current feature of the target object is less than the similarity threshold, based on not conflicting with the target object and with any other target The reference feature corresponding to the target object of the object conflict is determined, and the reference feature matching the current feature of the target object is determined.
在一种可能的实现方式中，所述装置还包括：In a possible implementation, the apparatus further includes:
第八确定部分，被配置为对于任一目标对象，响应于该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突的情况，将所述上一时刻不与其他目标对象冲突、且未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；或者，响应于该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况，将该目标对象在所述上一时刻所属的冲突区中、未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；The eighth determination part is configured to, for any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, the last moment does not conflict with other target objects. The target object that conflicts with the target object and does not match with other target objects at the current moment is determined as the remaining target object at the previous moment; or, in response to the target object at the current moment and the previous moment In the case of conflict with the same other target object, the target object in the conflict zone to which the target object belongs at the last moment and which does not match with other target objects at the current moment is determined as the remaining target object at the last moment. target;
第九确定部分，被配置为从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息。The ninth determination part is configured to, from the remaining target objects at the last moment, determine the target object with the closest distance to the target object at the current moment as the target matching the target object at the current moment object, and use the identification information of the matched target object as the identification information of the target object.
在本公开实施例以及其他的实施例中，“部分”可以是部分电路、部分处理器、部分程序或软件等等，当然也可以是单元，还可以是模块也可以是非模块化的。In the embodiments of the present disclosure and other embodiments, a "part" may be a part of a circuit, a part of a processor, a part of a program or software, etc., of course, a unit, a module or a non-modularity.
在本公开实施例中，通过根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，并将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息，由此能够保持目标对象的标识信息的一致性，即，可以保持同一目标对象在进出冲突区时标识信息的一致性，从而能够提高多目标跟踪的准确性。In the embodiment of the present disclosure, by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other target objects in the plurality of target objects When the target objects no longer conflict, extract the current feature of the target object, determine the reference feature matching the current feature of the target object based on the plurality of reference features corresponding to the plurality of target objects, and compare the reference feature with the target object's current features. The identification information corresponding to the reference feature of the current feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained , which can improve the accuracy of multi-target tracking.
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述。In some embodiments, the functions or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the methods described in the above method embodiments, and the implementation of the above method embodiments may refer to the descriptions of the above method embodiments. Repeat.
本公开实施例还提供一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。其中，所述计算机可读存储介质可以是非易失性计算机可读存储介质，或者可以是易失性计算机可读存储介质。Embodiments of the present disclosure further provide a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented. Wherein, the computer-readable storage medium may be a non-volatile computer-readable storage medium, or may be a volatile computer-readable storage medium.
本公开实施例还提供了一种计算机程序产品，包括计算机可读代码，当计算机可读代码在设备上运行时，设备中的处理器执行用于实现如上任一实施例提供的目标跟踪方法的指令。Embodiments of the present disclosure also provide a computer program product, including computer-readable codes. When the computer-readable codes are run on a device, a processor in the device executes a method for implementing the target tracking method provided in any of the above embodiments. instruction.
本公开实施例还提供了另一种计算机程序产品，用于存储计算机可读指令，指令被执行时使得计算机执行上述任一实施例提供的目标跟踪方法的操作。Embodiments of the present disclosure further provide another computer program product for storing computer-readable instructions, which, when executed, cause the computer to perform the operations of the target tracking method provided by any of the foregoing embodiments.
本公开实施例还提供一种电子设备，包括：一个或多个处理器；用于存储可执行指令的存储器；其中，所述一个或多个处理器被配置为调用所述存储器存储的可执行指令，以执行上述方法。Embodiments of the present disclosure further provide an electronic device, including: one or more processors; a memory for storing executable instructions; wherein the one or more processors are configured to invoke executable instructions stored in the memory instruction to execute the above method.
电子设备可以被提供为终端、服务器或其它形态的设备。The electronic device may be provided as a terminal, server or other form of device.
图7示出本公开实施例提供的一种电子设备800的框图。例如，电子设备800可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。FIG. 7 shows a block diagram of an electronic device 800 provided by an embodiment of the present disclosure. For example, electronic device 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, etc. terminal.
参照图7，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。7, an electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, a sensor component 814 , and the communication component 816 .
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。 例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 can include one or more processors 820 to execute instructions to perform all or some of the steps of the methods described above. Additionally, processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。 Memory 804 is configured to store various types of data to support operation at electronic device 800 . Examples of such data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, and the like. Memory 804 may be implemented by any type of volatile or nonvolatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。 Power supply assembly 806 provides power to various components of electronic device 800 . Power supply components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800 .
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。 Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action. In some embodiments, the multimedia component 808 includes a front-facing camera and/or a rear-facing camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。 Audio component 810 is configured to output and/or input audio signals. For example, audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when electronic device 800 is in operating modes, such as calling mode, recording mode, and voice recognition mode. The received audio signal may be further stored in memory 804 or transmitted via communication component 816 . In some embodiments, audio component 810 also includes a speaker for outputting audio signals.
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。The I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。 Sensor assembly 814 includes one or more sensors for providing status assessment of various aspects of electronic device 800 . For example, the sensor assembly 814 can detect the on/off state of the electronic device 800, the relative positioning of the components, such as the display and the keypad of the electronic device 800, the sensor assembly 814 can also detect the electronic device 800 or one of the electronic device 800 Changes in the position of components, presence or absence of user contact with the electronic device 800 , orientation or acceleration/deceleration of the electronic device 800 and changes in the temperature of the electronic device 800 . Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如Wi-Fi、2G、3G、4G/LTE、5G或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。 Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. The electronic device 800 may access wireless networks based on communication standards, such as Wi-Fi, 2G, 3G, 4G/LTE, 5G, or a combination thereof. In one exemplary embodiment, the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。In an exemplary embodiment, electronic device 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A programmed gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation is used to perform the above method.
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。In an exemplary embodiment, a non-volatile computer-readable storage medium, such as a memory 804 comprising computer program instructions executable by the processor 820 of the electronic device 800 to perform the above method is also provided.
图8示出本公开实施例提供的一种电子设备1900的框图。例如，电子设备1900可以被提供为一服务器。参照图8，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。FIG. 8 shows a block diagram of an electronic device 1900 provided by an embodiment of the present disclosure. For example, the electronic device 1900 may be provided as a server. 8, electronic device 1900 includes processing component 1922, which further includes one or more processors, and a memory resource represented by memory 1932 for storing instructions executable by processing component 1922, such as applications. An application program stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Additionally, the processing component 1922 is configured to execute instructions to perform the above-described methods.
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows
Figure PCTCN2021086020-appb-000004
Mac OS
Figure PCTCN2021086020-appb-000005
Figure PCTCN2021086020-appb-000006
或类似。 The electronic device 1900 may also include a power supply assembly 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input output (I/O) interface 1958 . Electronic device 1900 may operate based on an operating system stored in memory 1932, such as Windows
Figure PCTCN2021086020-appb-000004
Mac OS
Figure PCTCN2021086020-appb-000005
Figure PCTCN2021086020-appb-000006
or similar.
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as memory 1932 comprising computer program instructions executable by processing component 1922 of electronic device 1900 to perform the above-described method.
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。The present disclosure may be a system, method and/or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present disclosure.
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。A computer-readable storage medium may be a tangible device that can hold and store instructions for use by the instruction execution device. The computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (non-exhaustive list) of computer readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory), static random access memory (SRAM), portable compact disk read only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically coded devices, such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above. Computer-readable storage media, as used herein, are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, light pulses through fiber optic cables), or through electrical wires transmitted electrical signals.
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。The computer readable program instructions described herein may be downloaded to various computing/processing devices from a computer readable storage medium, or to an external computer or external storage device over a network such as the Internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer-readable program instructions from a network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。Computer program instructions for carrying out operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or instructions in one or more programming languages. Source or object code, written in any combination, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as the "C" language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider through the Internet connect). In some embodiments, custom electronic circuits, such as programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), can be personalized by utilizing state information of computer readable program instructions. Computer readable program instructions are executed to implement various aspects of the present disclosure.
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。Aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer readable program instructions.
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processor of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams. These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium on which the instructions are stored includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。Computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more functions for implementing the specified logical function(s) executable instructions. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.
该计算机程序产品可以具体通过硬件、软件或其结合的方式实现。在一个可选实施例中，所述计算机程序产品具体体现为计算机存储介质，在另一个可选实施例中，计算机程序产品具体体现为软件产品，例如软件开发包(Software Development Kit，SDK)等等。The computer program product can be specifically implemented by hardware, software or a combination thereof. In an optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), etc. Wait.
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。Various embodiments of the present disclosure have been described above, and the foregoing descriptions are exemplary, not exhaustive, and not limiting of the disclosed embodiments. Numerous modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to best explain the principles of the various embodiments, the practical application or improvement over the technology in the marketplace, or to enable others of ordinary skill in the art to understand the various embodiments disclosed herein.
工业实用性Industrial Applicability
本公开实施例中，通过根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象，响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，并将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息，由此能够保持目标对象的标识信息的一致性，即，可以保持同一目标对象在进出冲突区时标识信息的一致性，从而能够提高多目标跟踪的准确性。In the embodiment of the present disclosure, by determining a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area, in response to any target object in the plurality of target objects and other targets in the plurality of target objects When the objects are no longer in conflict, the current feature of the target object is extracted, and based on the multiple reference features corresponding to the multiple target objects, the reference feature matching the current feature of the target object is determined, and the reference feature matching the current feature of the target object is determined. The identification information corresponding to the reference feature of the feature matching is determined as the identification information of the target object, so that the consistency of the identification information of the target object can be maintained, that is, the consistency of the identification information of the same target object when entering and leaving the conflict area can be maintained, Thus, the accuracy of multi-target tracking can be improved.

Claims (25)

  1. 一种目标跟踪方法，包括：A target tracking method, comprising:
    根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象；According to the distance between the target objects in the tracking area, determine multiple target objects that conflict with each other;
    响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征；In response to a situation in which any one of the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects, extracting the current feature of the target object;
    基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目标对象提取的特征；Based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object is determined, wherein the plurality of reference features are the corresponding reference features of the target objects before the plurality of target objects collide with each other. Describe the features extracted from multiple target objects;
    将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。The identification information corresponding to the reference feature matching the current feature of the target object is determined as the identification information of the target object.
  2. 根据权利要求1所述的方法，其中，彼此冲突的多个目标对象中的任一目标对象，与所述多个目标对象中的其他至少一个目标对象之间的距离小于或等于距离阈值；The method according to claim 1, wherein a distance between any target object in the plurality of target objects that conflict with each other and at least one other target object in the plurality of target objects is less than or equal to a distance threshold;
    所述响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征，包括：The extracting the current feature of the target object in response to the situation that any target object in the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects includes:
    对于所述多个目标对象中的任一目标对象，响应于该目标对象与所述多个目标对象中的其他目标对象之间的距离均大于所述距离阈值的情况，提取该目标对象的当前特征。For any target object in the plurality of target objects, in response to the situation that the distance between the target object and other target objects in the plurality of target objects is greater than the distance threshold, extract the current value of the target object feature.
  3. 根据权利要求1或2所述的方法，其中，所述方法还包括：The method according to claim 1 or 2, wherein the method further comprises:
    对于任一目标对象，确定该目标对象对应于多个摄像头的第一坐标；其中，该目标对象对应于任一摄像头的第一坐标用于表征根据该摄像头采集的图像得到的该目标对象的第一坐标；For any target object, determine the first coordinates of the target object corresponding to multiple cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the first coordinates of the target object obtained according to the images collected by the camera. a coordinate;
    根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，其中，所述跟踪区域包括多个子区域；According to the first coordinates of the target object corresponding to the plurality of cameras, determine the sub-area where the target object is located in the tracking area, wherein the tracking area includes a plurality of sub-areas;
    根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头的第一坐标进行融合，得到该目标对象的融合坐标；According to the confidence of the plurality of cameras with respect to the sub-region where the target object is located, the first coordinates of the target object corresponding to the plurality of cameras are fused to obtain the fusion coordinates of the target object;
    在所述根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象之前，还包括：Before determining the multiple conflicting target objects according to the distances between the target objects in the tracking area, the method further includes:
    根据所述跟踪区域中目标对象的融合坐标，确定所述跟踪区域中目标对象之间的距离。The distance between the target objects in the tracking area is determined according to the fusion coordinates of the target objects in the tracking area.
  4. 根据权利要求1至3中任意一项所述的方法，其中，所述提取该目标对象的当前特征，包括：The method according to any one of claims 1 to 3, wherein the extracting the current feature of the target object comprises:
    确定该目标对象在所述跟踪区域中所处的子区域；determining the sub-area where the target object is located in the tracking area;
    根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头；Determine the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located;
    根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征。Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
  5. 根据权利要求4所述的方法，其中，所述根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头，包括：The method according to claim 4, wherein determining the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located, comprising:
    根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头。According to the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the The camera that extracts the current features of this target object.
  6. 根据权利要求5所述的方法，其中，所述重叠信息包括该目标对象的检测框与其他目标对象的检测框的交并比；The method according to claim 5, wherein the overlapping information includes the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
    所述根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头，包括：The determination is based on the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras. The camera used to extract the current features of the target object, including:
    从满足重叠条件的摄像头中，将针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，其中，所述重叠条件表示在所述当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值。From the cameras that satisfy the overlapping condition, the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition The intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
  7. 根据权利要求3至6中任意一项所述的方法，其中，所述方法还包括：The method according to any one of claims 3 to 6, wherein the method further comprises:
    对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度。For any camera in the plurality of cameras, according to the distance between the camera and the plurality of sub-areas of the tracking area, the confidence level of the camera for the plurality of sub-areas is determined.
  8. 根据权利要求7所述的方法，其中，在所述确定所述多个摄像头针对所述多个子区域的置信度之后，所述方法还包括：The method according to claim 7, wherein after the determining the confidence of the plurality of cameras with respect to the plurality of sub-regions, the method further comprises:
    对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，调整该摄像头针对所述多个子区域的置信度。For any camera in the plurality of cameras, according to the average distance between the target objects in the video frames collected by the camera, the confidence level of the camera with respect to the plurality of sub-regions is adjusted.
  9. 根据权利要求1至8中任意一项所述的方法，其中，所述基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，包括：The method according to any one of claims 1 to 8, wherein the determining, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature matching the current feature of the target object comprises:
    从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征；From the plurality of reference features corresponding to the plurality of target objects, determine the reference feature with the greatest similarity to the current feature of the target object;
    响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。In response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold, determine the reference feature with the greatest similarity as the reference matching the current feature of the target object feature.
  10. 根据权利要求9所述的方法，其中，在所述确定与该目标对象的当前特征的相似度最大的基准特征之后，所述方法还包括：The method according to claim 9, wherein, after said determining the reference feature with the greatest similarity with the current feature of the target object, the method further comprises:
    响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征。In response to the situation where the similarity between the reference feature with the greatest similarity and the current feature of the target object is less than the similarity threshold, based on the reference corresponding to the target object that does not conflict with the target object and conflicts with any other target object feature to determine the reference feature that matches the current feature of the target object.
  11. 根据权利要求1至10中任意一项所述的方法，其中，所述方法还包括：The method according to any one of claims 1 to 10, wherein the method further comprises:
    对于任一目标对象，响应于该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突的情况，将所述上一时刻不与其他目标对象冲突、且未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；或者，响应于该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况，将该目标对象在所述上一时刻所属的冲突区中、未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；For any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, the last moment does not conflict with other target objects and does not conflict with all target objects. The target object matched by other target objects at the current moment is determined to be the remaining target objects at the previous moment; situation, the target object in the conflict zone to which the target object belongs at the last moment, and the target object that does not match with other target objects at the current moment, is determined as the remaining target object at the last moment;
    从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息。From the remaining target objects at the last moment, the target object with the closest distance to the target object at the current moment is determined as the target object matching the target object at the current moment, and the matching target The identification information of the object is used as the identification information of the target object.
  12. 一种目标跟踪装置，包括：A target tracking device, comprising:
    第一确定部分，被配置为根据跟踪区域中目标对象之间的距离，确定彼此冲突的多个目标对象；a first determining part, configured to determine a plurality of target objects that conflict with each other according to the distance between the target objects in the tracking area;
    提取部分，被配置为响应于所述多个目标对象中的任一目标对象与多个目标对象中的其他目标对象不再冲突的情况，提取该目标对象的当前特征；an extraction part, configured to extract the current feature of the target object in response to a situation in which any one of the plurality of target objects is no longer in conflict with other target objects in the plurality of target objects;
    第二确定部分，被配置为基于所述多个目标对象对应的多个基准特征，确定与该目标对象的当前特征匹配的基准特征，其中，所述多个基准特征是在所述多个目标对象彼此冲突之前，分别对所述多个目标对象提取的特征；The second determination part is configured to determine, based on a plurality of reference features corresponding to the plurality of target objects, a reference feature that matches the current feature of the target object, wherein the plurality of reference features are in the plurality of targets Before the objects collide with each other, the features extracted from the multiple target objects respectively;
    第三确定部分，被配置为将与该目标对象的当前特征匹配的基准特征对应的标识信息，确定为该目标对象的标识信息。The third determining part is configured to determine, as the identification information of the target object, the identification information corresponding to the reference feature matched with the current feature of the target object.
  13. 根据权利要求12所述的装置，其中，所述提取部分被配置为：13. The apparatus of claim 12, wherein the extraction portion is configured to:
    对于所述多个目标对象中的任一目标对象，响应于该目标对象与所述多个目标对象中的其他目标对象之间的距离均大于所述距离阈值的情况，提取该目标对象的当前特征。For any target object in the plurality of target objects, in response to the situation that the distance between the target object and other target objects in the plurality of target objects is greater than the distance threshold, extract the current value of the target object feature.
  14. 根据权利要求12或13所述的装置，其中，所述装置还包括：The apparatus of claim 12 or 13, wherein the apparatus further comprises:
    第三确定部分，被配置为对于任一目标对象，确定该目标对象对应于多个摄像头的第一坐标；其中，该目标对象对应于任一摄像头的第一坐标用于表征根据该摄像头采集的图像得到的该目标对象的第一坐标；The third determining part is configured to, for any target object, determine the first coordinates of the target object corresponding to the plurality of cameras; wherein, the first coordinates of the target object corresponding to any camera are used to represent the data collected according to the camera. the first coordinate of the target object obtained from the image;
    第四确定部分，被配置为根据该目标对象对应于所述多个摄像头的第一坐标，确定该目标对象在所述跟踪区域中所处的子区域，其中，所述跟踪区域包括多个子区域；a fourth determining part, configured to determine a sub-area where the target object is located in the tracking area according to the first coordinates of the target object corresponding to the plurality of cameras, wherein the tracking area includes a plurality of sub-areas ;
    融合部分，被配置为根据所述多个摄像头针对该目标对象所处的子区域的置信度，对该目标对象对应于所述多个摄像头的第一坐标进行融合，得到该目标对象的融合坐标；The fusion part is configured to fuse the first coordinates of the target object corresponding to the plurality of cameras according to the confidence of the plurality of cameras for the sub-region where the target object is located to obtain the fusion coordinates of the target object ;
    第五确定部分，被配置为根据所述跟踪区域中目标对象的融合坐标，确定所述跟踪区域中目标对象之间的距离。The fifth determination part is configured to determine the distance between the target objects in the tracking area according to the fusion coordinates of the target objects in the tracking area.
  15. 根据权利要求12-14任一项所述的装置，其中，所述提取部分被配置为：The apparatus of any of claims 12-14, wherein the extraction portion is configured to:
    确定该目标对象在所述跟踪区域中所处的子区域；determining the sub-area where the target object is located in the tracking area;
    根据多个摄像头针对该目标对象所处的子区域的置信度，确定用于提取该目标对象的当前特征的摄像头；Determine the camera used to extract the current feature of the target object according to the confidence of the plurality of cameras with respect to the sub-region where the target object is located;
    根据用于提取该目标对象的当前特征的摄像头采集的视频帧，提取该目标对象的当前特征。Extract the current feature of the target object according to the video frame collected by the camera used for extracting the current feature of the target object.
  16. 根据权利要求15所述的装置，其中，所述提取部分被配置为：16. The apparatus of claim 15, wherein the extraction portion is configured to:
    根据多个摄像头针对该目标对象所处的子区域的置信度，以及所述多个摄像头采集的视频帧中该目标对象的检测框与其他目标对象的检测框之间的重叠信息，确定用于提取该目标对象的当前特征的摄像头。According to the confidence of the multiple cameras for the sub-region where the target object is located, and the overlap information between the detection frame of the target object and the detection frames of other target objects in the video frames collected by the multiple cameras, determine the The camera that extracts the current features of this target object.
  17. 根据权利要求16所述的装置，其中，所述重叠信息包括该目标对象的检测框与其他目标对象的检测框的交并比；The device according to claim 16, wherein the overlapping information comprises the intersection ratio of the detection frame of the target object and the detection frames of other target objects;
    所述提取部分被配置为：The extraction section is configured as:
    从满足重叠条件的摄像头中，将针对该目标对象所处的子区域的置信度最高的摄像头确定为用于提取该目标对象的当前特征的摄像头，其中，所述重叠条件表示在所述当前采集的视频帧中该目标对象的检测框与其他目标对象的检测框的交并比均小于预定阈值。From the cameras that satisfy the overlapping condition, the camera with the highest confidence for the sub-region where the target object is located is determined as the camera for extracting the current feature of the target object, wherein the overlapping condition indicates that the current acquisition The intersection ratios of the detection frame of the target object and the detection frames of other target objects in the video frame of the target object are all smaller than a predetermined threshold.
  18. 根据权利要求14-17任一项所述的装置，其中，所述装置还包括：The apparatus of any one of claims 14-17, wherein the apparatus further comprises:
    第六确定部分，被配置为对于所述多个摄像头中的任一摄像头，根据该摄像头与所述跟踪区域的多个子区域之间的距离，确定该摄像头针对所述多个子区域的置信度。The sixth determining part is configured to, for any one of the plurality of cameras, determine the confidence of the camera for the plurality of sub-areas according to the distance between the camera and the plurality of sub-areas of the tracking area.
  19. 根据权利要求18所述的装置，其中，所述装置还包括：The apparatus of claim 18, wherein the apparatus further comprises:
    调整部分，被配置为对于所述多个摄像头中的任一摄像头，根据该摄像头采集的视频帧中目标对象之间的平均距离，调整该摄像头针对所述多个子区域的置信度。The adjusting part is configured to, for any one of the plurality of cameras, adjust the confidence of the camera for the plurality of sub-regions according to the average distance between the target objects in the video frames collected by the camera.
  20. 根据权利要求12-18任一项所述的装置，其中，所述第二确定部分被配置为：The apparatus of any one of claims 12-18, wherein the second determining portion is configured to:
    从所述多个目标对象对应的多个基准特征中，确定与该目标对象的当前特征的相似度最大的基准特征；From the plurality of reference features corresponding to the plurality of target objects, determine the reference feature with the greatest similarity to the current feature of the target object;
    响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度大于或等于相似度阈值的情况，将该相似度最大的基准特征确定为与该目标对象的当前特征匹配的基准特征。In response to the situation that the similarity between the reference feature with the greatest similarity and the current feature of the target object is greater than or equal to the similarity threshold, determine the reference feature with the greatest similarity as the reference matching the current feature of the target object feature.
  21. 根据权利要求20所述的装置，其中，所述装置还包括：The apparatus of claim 20, wherein the apparatus further comprises:
    第七确定部分，被配置为响应于该相似度最大的基准特征与该目标对象的当前特征之间的相似度小于相似度阈值的情况，基于不与该目标对象冲突、且与任一其他目标对象冲突的目标对象对应的基准特征，确定与该目标对象的当前特征匹配的基准特征。A seventh determination part, configured to respond to a situation where the similarity between the reference feature with the greatest similarity and the current feature of the target object is less than the similarity threshold, based on not conflicting with the target object and with any other target The reference feature corresponding to the target object of the object conflict is determined, and the reference feature matching the current feature of the target object is determined.
  22. 根据权利要求12-21任一项所述的装置，其中，所述装置还包括：The apparatus of any one of claims 12-21, wherein the apparatus further comprises:
    第八确定部分，被配置为对于任一目标对象，响应于该目标对象在当前时刻和所述当前时刻的上一时刻均不与其他目标对象冲突的情况，将所述上一时刻不与其他目标对象冲突、且未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；或者，响应于该目标对象在所述当前时刻和所述上一时刻与相同的其他目标对象冲突的情况，将该目标对象在所述上一时刻所属的冲突区中、未与所述当前时刻的其他目标对象匹配的目标对象，确定为所述上一时刻的剩余目标对象；The eighth determination part is configured to, for any target object, in response to the situation that the target object does not conflict with other target objects at the current moment and the previous moment of the current moment, the last moment does not conflict with other target objects. The target object that conflicts with the target object and does not match with other target objects at the current moment is determined as the remaining target object at the previous moment; or, in response to the target object at the current moment and the previous moment In the case of conflict with the same other target object, the target object in the conflict zone to which the target object belongs at the last moment and which does not match with other target objects at the current moment is determined as the remaining target object at the last moment. target;
    第九确定部分，被配置为从所述上一时刻的剩余目标对象中，将与所述当前时刻的该目标对象距离最近的目标对象，确定为与所述当前时刻的该目标对象匹配的目标对象，并将该匹配的目标对象的标识信息作为该目标对象的标识信息。The ninth determination part is configured to, from the remaining target objects at the last moment, determine the target object with the closest distance to the target object at the current moment as the target matching the target object at the current moment object, and use the identification information of the matched target object as the identification information of the target object.
  23. 一种电子设备，包括：An electronic device comprising:
    一个或多个处理器；one or more processors;
    用于存储可执行指令的存储器；memory for storing executable instructions;
    其中，所述一个或多个处理器被配置为调用所述存储器存储的可执行指令，以执行权利要求1至11中任意一项所述的方法。wherein the one or more processors are configured to invoke executable instructions stored in the memory to perform the method of any one of claims 1-11.
  24. 一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现权利要求1至11中任意一项所述的方法。A computer-readable storage medium having computer program instructions stored thereon, the computer program instructions implementing the method of any one of claims 1 to 11 when executed by a processor.
  25. 一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行时实现权利要求1至11中任意一项所述的方法。A computer program comprising computer readable code, when the computer readable code is executed in an electronic device, the processor in the electronic device implements the method of any one of claims 1 to 11 when executed.
PCT/CN2021/086020 2020-11-11 2021-04-08 Object tracking method and apparatus, electronic device, and storage medium WO2022099988A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2022506434A JP2022552772A (en) 2020-11-11 2021-04-08 Target tracking method and apparatus, electronic device and storage medium
KR1020227002278A KR102446688B1 (en) 2020-11-11 2021-04-08 Target tracking method and apparatus, electronic device and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011256801.6A CN112330717B (en) 2020-11-11 2020-11-11 Target tracking method and device, electronic equipment and storage medium
CN202011256801.6 2020-11-11

Publications (1)

Publication Number Publication Date
WO2022099988A1 true WO2022099988A1 (en) 2022-05-19

Family

ID=74318993

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/086020 WO2022099988A1 (en) 2020-11-11 2021-04-08 Object tracking method and apparatus, electronic device, and storage medium

Country Status (4)

Country Link
JP (1) JP2022552772A (en)
KR (1) KR102446688B1 (en)
CN (1) CN112330717B (en)
WO (1) WO2022099988A1 (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112330717B (en) * 2020-11-11 2023-03-10 北京市商汤科技开发有限公司 Target tracking method and device, electronic equipment and storage medium
CN113129339B (en) * 2021-04-28 2023-03-10 北京市商汤科技开发有限公司 Target tracking method and device, electronic equipment and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107831490A (en) * 2017-12-01 2018-03-23 南京理工大学 A kind of improved more extension method for tracking target
US20190114788A1 (en) * 2016-07-08 2019-04-18 Omron Corporation Image processing device and image processing method
CN110827325A (en) * 2019-11-13 2020-02-21 北京百度网讯科技有限公司 Target tracking method and device, electronic equipment and storage medium
CN111369590A (en) * 2020-02-27 2020-07-03 北京三快在线科技有限公司 Multi-target tracking method and device, storage medium and electronic equipment
CN111860373A (en) * 2020-07-24 2020-10-30 浙江商汤科技开发有限公司 Target detection method and device, electronic equipment and storage medium
CN112330717A (en) * 2020-11-11 2021-02-05 北京市商汤科技开发有限公司 Target tracking method and device, electronic equipment and storage medium

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002342762A (en) * 2001-05-22 2002-11-29 Matsushita Electric Ind Co Ltd Object tracing method
KR101355974B1 (en) * 2010-08-24 2014-01-29 한국전자통신연구원 Method and devices for tracking multiple object
US9147260B2 (en) * 2010-12-20 2015-09-29 International Business Machines Corporation Detection and tracking of moving objects
KR101406334B1 (en) * 2013-04-18 2014-06-19 전북대학교산학협력단 System and method for tracking multiple object using reliability and delayed decision
KR101868103B1 (en) * 2017-07-12 2018-06-18 군산대학교 산학협력단 A video surveillance apparatus for identification and tracking multiple moving objects and method thereof
CN111275737B (en) * 2020-01-14 2023-09-12 北京市商汤科技开发有限公司 Target tracking method, device, equipment and storage medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190114788A1 (en) * 2016-07-08 2019-04-18 Omron Corporation Image processing device and image processing method
CN107831490A (en) * 2017-12-01 2018-03-23 南京理工大学 A kind of improved more extension method for tracking target
CN110827325A (en) * 2019-11-13 2020-02-21 北京百度网讯科技有限公司 Target tracking method and device, electronic equipment and storage medium
CN111369590A (en) * 2020-02-27 2020-07-03 北京三快在线科技有限公司 Multi-target tracking method and device, storage medium and electronic equipment
CN111860373A (en) * 2020-07-24 2020-10-30 浙江商汤科技开发有限公司 Target detection method and device, electronic equipment and storage medium
CN112330717A (en) * 2020-11-11 2021-02-05 北京市商汤科技开发有限公司 Target tracking method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
KR102446688B1 (en) 2022-09-23
KR20220066245A (en) 2022-05-24
CN112330717A (en) 2021-02-05
JP2022552772A (en) 2022-12-20
CN112330717B (en) 2023-03-10

Similar Documents

Publication Publication Date Title
US9953506B2 (en) Alarming method and device
TWI766286B (en) Image processing method and image processing device, electronic device and computer-readable storage medium
CN106651955B (en) Method and device for positioning target object in picture
JP6392468B2 (en) Region recognition method and apparatus
CN108629354B (en) Target detection method and device
TW202131219A (en) Image recognition method and apparatus, electronic device, and storage medium
US10007841B2 (en) Human face recognition method, apparatus and terminal
CN108010060B (en) Target detection method and device
TWI702544B (en) Method, electronic device for image processing and computer readable storage medium thereof
TW202029055A (en) Pedestrian recognition method and device
CN107480665B (en) Character detection method and device and computer readable storage medium
CN107784279B (en) Target tracking method and device
CN106557759B (en) Signpost information acquisition method and device
CN111104920B (en) Video processing method and device, electronic equipment and storage medium
WO2022099988A1 (en) Object tracking method and apparatus, electronic device, and storage medium
TW202219833A (en) Detection method for pedestrian events, electronic device, and storage medium
US11574415B2 (en) Method and apparatus for determining an icon position
CN109344703B (en) Object detection method and device, electronic equipment and storage medium
CN112906484A (en) Video frame processing method and device, electronic equipment and storage medium
CN111680646A (en) Motion detection method and device, electronic device and storage medium
US20230048952A1 (en) Image registration method and electronic device
WO2023155350A1 (en) Crowd positioning method and apparatus, electronic device, and storage medium
US20220383517A1 (en) Method and device for target tracking, and storage medium
CN112330721B (en) Three-dimensional coordinate recovery method and device, electronic equipment and storage medium
CN114863392A (en) Lane line detection method, lane line detection device, vehicle, and storage medium

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022506434

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21890523

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21890523

Country of ref document: EP

Kind code of ref document: A1