WO2019042196A1 - 一种人体目标轨迹确定方法及装置 - Google Patents

一种人体目标轨迹确定方法及装置 Download PDF

Info

Publication number
WO2019042196A1
WO2019042196A1 PCT/CN2018/101666 CN2018101666W WO2019042196A1 WO 2019042196 A1 WO2019042196 A1 WO 2019042196A1 CN 2018101666 W CN2018101666 W CN 2018101666W WO 2019042196 A1 WO2019042196 A1 WO 2019042196A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
feature
target feature
collection
attribute
Prior art date
Application number
PCT/CN2018/101666
Other languages
English (en)
French (fr)
Inventor
浦世亮
申琳
沈林杰
Original Assignee
杭州海康威视数字技术股份有限公司
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 杭州海康威视数字技术股份有限公司 filed Critical 杭州海康威视数字技术股份有限公司
Priority to US16/643,373 priority Critical patent/US11157720B2/en
Priority to EP18852621.4A priority patent/EP3678057B1/en
Publication of WO2019042196A1 publication Critical patent/WO2019042196A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30196—Human being; Person
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30241—Trajectory

Definitions

  • the present application relates to the field of image processing technologies, and in particular, to a method and apparatus for determining a human target trajectory.
  • the face grabbing machine is usually used to track the human trajectory: a face grabbing machine is set at various positions such as roads and public places, and the face image collected by the face grabbing machine determines the time and place where a specific person appears, thereby The trajectory of this particular person is tracked.
  • the purpose of the embodiments of the present application is to provide a method and a device for determining a human target trajectory to improve the accuracy of determining a human trajectory.
  • the embodiment of the present application provides a method for determining a human target trajectory, including: acquiring an image to be processed; and extracting a target feature of the human target to be tracked in the image to be processed as a target feature to be searched; The corresponding relationship between the target feature and the collected attribute is used to find the collection attribute corresponding to the target feature to be searched; wherein the collection attribute corresponding to a target feature in the corresponding relationship is an collection attribute of the image having the target feature, The collection attribute includes a collection location; and the trajectory of the human target to be tracked is determined according to the found collection attribute.
  • the method further includes: extracting a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched; and based on the pre-established facial feature and Collecting the corresponding relationship of the attributes, and searching for the collection attribute corresponding to the face feature to be searched; wherein the collection attribute corresponding to a face feature in the correspondence relationship is an collection attribute of the image having the face feature.
  • the method further includes: extracting a facial feature of the human target to be tracked in the image to be processed as a facial feature to be searched; and the pre-established target feature
  • the step of searching for the collection attribute corresponding to the target feature to be searched may include: searching for the target feature to be searched based on the corresponding relationship between the target feature and the face feature and the collected attribute.
  • the collection attribute of the face feature to be searched is used as the collection attribute corresponding to the target feature to be searched; wherein a corresponding target feature and a face feature belong to the same human target in the corresponding relationship;
  • the acquisition attribute corresponding to a target feature and the face feature in the relationship is an acquisition attribute of the image having the target feature and the face feature.
  • the step of searching for the collection attribute corresponding to the hash value to be searched according to the correspondence between the pre-established hash value and the collection attribute may include: separately calculating the pre-established hash value and the collection attribute. a similarity between each hash value included in the correspondence relationship and the hash value to be searched; and an acquisition attribute corresponding to the hash value whose similarity satisfies the preset condition is determined.
  • the method further includes: determining an collection attribute of the image to be processed as a collection attribute to be searched; and determining, according to the pre-established correspondence between the target feature and the collection attribute,
  • the step of searching for the collection attribute corresponding to the target feature to be searched may include: searching, in a correspondence between the target feature and the collection attribute, the target collection attribute whose difference from the to-be-searched collection attribute is less than a preset threshold. And the target feature corresponding to the target collection attribute, as the target feature to be matched; determining whether the target feature to be matched matches the target feature to be searched; if yes, using the target collection attribute as the target to be searched The collection attribute corresponding to the feature.
  • the step of searching for the collection attribute corresponding to the target feature to be searched based on the correspondence between the pre-established target feature and the collection attribute may include: in a correspondence between the pre-established target feature and the collection attribute, Searching for a target feature set that matches the target feature to be searched; wherein the target feature set is composed of each target feature belonging to the same human target; and the collection attribute corresponding to each target feature included in the target feature set As the collection attribute corresponding to the target feature to be searched.
  • the collection attribute further includes an acquisition moment.
  • the embodiment of the present application further provides a human target trajectory determining apparatus, including: an acquiring module, configured to acquire an image to be processed; and a first extracting module, configured to extract a human target to be tracked in the image to be processed
  • the target feature is used as the target feature to be searched;
  • the first search module is configured to search for the collection attribute corresponding to the target feature to be searched based on the corresponding relationship between the target feature and the collected attribute; wherein, the corresponding relationship is
  • the collection attribute corresponding to the target feature is an acquisition attribute of the image having the target feature, the collection attribute includes a collection location, and the first determining module is configured to determine the trajectory of the target to be tracked according to the found collection attribute.
  • the device may further include: a second extracting module, configured to extract a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched; and a second searching module, configured to Corresponding relationship between the created face feature and the collected attribute, and searching for the collected attribute corresponding to the face feature to be searched; wherein, the corresponding attribute of the face feature in the corresponding relationship is an image having the face feature Collect properties.
  • a second extracting module configured to extract a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched
  • a second searching module configured to Corresponding relationship between the created face feature and the collected attribute, and searching for the collected attribute corresponding to the face feature to be searched; wherein, the corresponding attribute of the face feature in the corresponding relationship is an image having the face feature Collect properties.
  • the device may further include: a third extracting module, configured to extract a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched;
  • the method is configured to: search for an acquisition attribute that matches the to-be-find target feature and the to-be-finished face feature, as a corresponding to the target feature to be searched, based on the correspondence between the target feature and the face feature, and the collection attribute.
  • the collection attribute wherein a corresponding target feature and the face feature belong to the same human target in the correspondence relationship; and an acquisition attribute corresponding to a target feature and a face feature in the correspondence relationship has the target feature and the person
  • the collection properties of the image of the face feature is configured to: search for an acquisition attribute that matches the to-be-find target feature and the to-be-finished face feature, as a corresponding to the target feature to be searched, based on the correspondence between the target feature and the face feature, and the collection attribute.
  • the collection attribute wherein a corresponding target feature and the face feature belong
  • the first extraction module may be configured to: extract an original target feature of the human target to be tracked in the image to be processed, and calculate a hash value of the original target feature as a hash value to be searched;
  • the first searching module may be configured to: search for an collection attribute corresponding to the hash value to be searched based on a correspondence between the pre-established hash value and the collection attribute.
  • the first searching module may be specifically configured to separately calculate a similarity between each hash value included in the correspondence between the pre-established hash value and the collection attribute and the hash value to be searched. Degree; determines an acquisition attribute corresponding to a hash value whose similarity satisfies a preset condition.
  • the device may further include: a second determining module, configured to determine an collection attribute of the to-be-processed image, as a collection attribute to be searched; the first searching module may be specifically configured to: pre-established In the corresponding relationship between the target feature and the collection attribute, the target acquisition attribute that is smaller than the preset threshold value and the target feature corresponding to the target collection attribute are searched as the target feature to be matched; Whether the target feature to be matched matches the target feature to be searched; if yes, the target acquisition attribute is used as the collection attribute corresponding to the target feature to be searched.
  • a second determining module configured to determine an collection attribute of the to-be-processed image, as a collection attribute to be searched
  • the first searching module may be specifically configured to: pre-established In the corresponding relationship between the target feature and the collection attribute, the target acquisition attribute that is smaller than the preset threshold value and the target feature corresponding to the target collection attribute are searched as the target feature to be matched; Whether the target feature to be matched matches the target feature to be
  • the first searching module may be configured to: in a correspondence between the pre-established target feature and the collected attribute, search for a target feature group that matches the target feature to be searched; wherein the target feature The group consists of each of the target features belonging to the same human target; the collection attribute corresponding to each target feature included in the target feature set is used as the collection attribute corresponding to the target feature to be searched.
  • the collection attribute further includes an acquisition moment.
  • an embodiment of the present application further provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through a communication bus;
  • the computer program is stored; and the processor is configured to implement any one of the above-mentioned human target trajectory determining methods when executing the program stored in the memory.
  • an embodiment of the present application further provides a computer readable storage medium, where the computer readable storage medium stores a computer program, and when the computer program is executed by the processor, implement any one of the above-mentioned human target trajectories. Determine the method.
  • an embodiment of the present application further provides an executable program code for being executed to implement any of the above-described human target trajectory determining methods.
  • FIG. 1 is a schematic flowchart of a method for determining a human target trajectory according to an embodiment of the present application
  • FIG. 2 is a schematic diagram of an application scenario provided by an embodiment of the present application.
  • FIG. 3 is a schematic diagram of a trajectory determined in the application scenario of FIG. 2;
  • FIG. 4 is a schematic structural diagram of a human target trajectory determining apparatus according to an embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
  • the embodiment of the present application provides a method and an apparatus for determining a human target trajectory.
  • the method and device can be applied to a device having an image processing function, which is not limited.
  • FIG. 1 is a schematic flowchart of determining a human target trajectory according to an embodiment of the present application, including:
  • S101 Acquire an image to be processed.
  • S102 Extract a target feature of the human target to be tracked in the image to be processed as a target feature to be searched.
  • S103 Searching for an acquisition attribute corresponding to the target feature to be searched based on the correspondence between the target feature and the collection attribute, where the collection attribute corresponding to a target feature in the correspondence is an acquisition attribute of the image having the target feature , the collection attribute contains the collection location.
  • S104 Determine a trajectory of the human target to be tracked according to the found collection attribute.
  • S101 Acquire an image to be processed.
  • the image to be processed is an image containing a human target to be tracked.
  • S101 may include: receiving a to-be-processed image input by a user; or, as another implementation manner, S101 may include: acquiring an image to be processed from a specified collection device.
  • the user can input an image containing the human target; or, the acquisition device that collects the human target can be determined, and an image including the human target is acquired from the collection device.
  • the image to be processed may be obtained in other manners, which is not limited.
  • S102 Extract a target feature of the human target to be tracked in the image to be processed as a target feature to be searched.
  • the target features of the human target may include colors, textures, sizes, and the like, and may also include the characteristics of the worn clothing, such as whether the type of the backpack or the pants, or the height and shape of the human target, and are not limited. .
  • S103 Searching for an acquisition attribute corresponding to the target feature to be searched based on the correspondence between the target feature and the collection attribute, where the collection attribute corresponding to a target feature in the correspondence is an acquisition attribute of the image having the target feature , the collection attribute contains the collection location.
  • the collection attribute may only include the collection location; in the second case, the collection attribute may include the collection location and the collection time.
  • one or more acquisition devices can be communicatively coupled to a server that can transmit acquired images to the server in real time, and the server receives images and extracts targets of human targets in the images. Features, and sequentially store target features of each image according to the order of reception time.
  • the server can determine the collection location of each image received, and establish a correspondence between the collection location of the same image and the target feature.
  • the server determines the collection location of each image received, for example, pre-acquiring the collection location information of each device, and establishing a correspondence between the device identifier and the collection location.
  • the device identifies the corresponding collection location as the collection location of the image; or the collection device can send the collected image together with the self-collection location information to the server; etc., no longer enumerated.
  • the server first receives the image A1 sent by the collection device A, extracts the target feature A2 of the human target in A1, and determines the collection of the image A1.
  • the location is A3, and the corresponding relationship between the target feature A2 and the collection location A3 is established.
  • the server receives the image B1 sent by the collection device B, extracts the target feature B2 of the human target in B1, and determines that the collection location of the image B1 is B3.
  • the server further receives the image C1 sent by the collection device C, extracts the target feature C2 of the human target in C1, and determines that the collection location of the image C1 is C3, and establishes the target feature C2.
  • the correspondence established by the server can be as shown in Table 1:
  • Target feature Collection location C2 C3 B2 B3 A2 A3
  • the correspondences of the pairs are stored in the order of the receiving time from late to early, or the correspondences of the pairs may be stored in the order of the receiving time from the morning to the night, which is not limited.
  • the established correspondence (such as Table 1) can be stored in the server local database or stored in a database connected to the server.
  • the execution entity of the embodiment of the present application and the server may be the same device or different devices.
  • Table 1 is only a simple example, and does not constitute a limitation on the actually established correspondence.
  • the second case is exemplified: one or more collection devices can be communicatively connected to a server, and the collection devices can send the collected images to the server in real time or in non-real time; the server receives images and extracts The target feature of the human target in the image, and determine the collection location of each image received at the acquisition time, and establish the corresponding relationship between the acquisition time, the collection location and the target feature of the same image.
  • the server may use the time when the image is received as the collection time; or the collection device may send the collected image and the collection time of the image to the server.
  • the collection device sends the collected image to the server in real time, the collection device sends the collected image and the collection time of the image to the server.
  • the server determines the collection location of each image received, for example, pre-acquiring the collection location information of each device, and establishing a correspondence between the device identifier and the collection location.
  • the device identifies the corresponding collection location as the collection location of the image; or the collection device can send the collected image together with the self-collection location information to the server; etc., no longer enumerated.
  • the server receives the image A1 sent by the collection device A, extracts the target feature A2 of the human target in A1, and determines the location of the image A1. For A3, the acquisition time is A4, and the correspondence between the target feature A2 and the collection location A3 and the acquisition time A4 is established; the server receives the image B1 sent by the collection device B, extracts the target feature B2 of the human target in B1, and determines the image B1 of the image B1.
  • the collection location is B3, the acquisition time is B4, the correspondence between the target feature B2 and the collection location B3, and the acquisition time B4 is established;
  • the server receives the image C1 sent by the collection device C, extracts the target feature C2 of the human target in C1, and determines the image.
  • the C1 collection location is C3, the acquisition time is A4, and the correspondence between the target feature C2 and the collection location C3 and the acquisition time C4 is established; the correspondence established by the server can be as shown in Table 2:
  • Target feature Collection location Acquisition moment C2 C3 C4 B2 B3 B4 A2 A3 A4
  • the above established correspondence (such as Table 2) can be stored in the server local database or stored in a database connected to the server.
  • the execution entity of the embodiment of the present application and the server may be the same device or different devices.
  • Table 2 is only a simple example and does not constitute a limitation on the actually established correspondence.
  • S104 Determine a trajectory of the human target to be tracked according to the found collection attribute.
  • the acquisition attribute includes the collection time and the collection location, and the trajectory can be determined according to the time and place.
  • the scenario shown in Figure 3 is described: It is assumed that three acquisition attributes of the same human target are found in the established correspondence, respectively: 1. Acquisition time: 9:5 am on July 14, 2017, collection location: B3, 2, collection time: 9:3 am on July 14, 2017, collection location: C3, 3, collection time: 9:00 am on July 14, 2017, collection location: F3, you can determine the person
  • the trajectory is shown by the dotted line in Fig. 3, from F3 to C3 to B3.
  • the trajectory may be determined according to the collection location and the order.
  • the scenario shown in Figure 3 is described: It is assumed that three acquisition attributes of the same human target are found in the established correspondence, which are: 1, collection location: B3, 2, collection location: C3, 3, collection location: F3 . It is assumed that the storage order of the correspondence relationship is stored in the order of receiving time from late to early, that is, among the three collection attributes, the third item with the earliest reception time is the first one, so the first one can be determined.
  • the trajectory of the person is shown by the dotted line in Fig. 3, from F3 to C3 to B3.
  • a correspondence between a face feature and an acquisition attribute may also be established.
  • the method further includes: extracting a face feature of the target to be tracked in the image to be processed, as a Searching for a face feature; searching for an acquisition attribute corresponding to the face feature to be searched based on a correspondence between the pre-established face feature and the collection attribute; wherein the collection attribute corresponding to a face feature in the correspondence relationship has The collection property of the image of the face feature.
  • establishing a correspondence between the face feature and the collection attribute may include:
  • the process of establishing the correspondence between the face feature and the collection attribute is similar to the process of establishing the correspondence between the target feature and the collection attribute, and will not be described again.
  • the image to be processed includes not only the target feature of the human target but also the facial feature of the human target; and in addition to the correspondence between the target feature and the collected attribute, the correspondence between the face feature and the collected attribute is also established. Therefore, on the basis of the embodiment shown in FIG. 1, more collection attributes can be found according to the facial features, and more accurate trajectories are determined based on more collection attributes.
  • the correspondence between the target feature, the face feature, and the collection attribute may be established, and the specific process may include:
  • the method further includes: extracting a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched;
  • the S103 may include: searching for an acquisition attribute that matches the to-be-find target feature and the to-be-finished face feature, based on the correspondence between the pre-established target feature and the face feature, and the collection attribute, as the target feature to be searched Corresponding collection properties.
  • a matching rule can be set. For example, the similarity between the target feature in the correspondence relationship and the target feature to be searched is greater than the first preset threshold, and the similarity between the face feature and the face feature to be searched in the corresponding relationship is greater than a second preset threshold.
  • the preset threshold and the second preset threshold may be set according to actual conditions, and may be the same or different.
  • the specific matching rules are not limited.
  • the target features and facial features may be extracted for each target to be tracked as the target features to be searched and the features to be found.
  • the collection attribute matching the target feature to be searched and the face feature to be found is searched for. That is to say, when the collection attribute matching the to-be-searched target feature and the to-be-find facial feature is searched, it is reasonable that the to-be-find target feature and the to-be-find facial feature belong to the same human target.
  • searching for the collection attribute matching the target feature and the face feature can improve the accuracy of the search.
  • the target feature of the human target and the facial feature may be represented by a hash value.
  • the target feature extracted in S102 is a hash value, and the target feature in the pre-established correspondence relationship is also a hash value.
  • the S103 may include: extracting an original target feature of the human target to be recognized in the image to be processed, and calculating a hash value of the original target feature as a hash value to be searched.
  • the extracting the facial features may include: extracting original facial features of the human target to be recognized in the image to be processed, and calculating a hash value of the original facial features as a hash value to be searched.
  • the S104 or the locating attribute corresponding to the face feature may include: searching for the collection attribute corresponding to the hash value to be searched based on the correspondence between the pre-established hash value and the collected attribute.
  • the hash value is used to represent the target feature and the face feature, and the search efficiency can be improved.
  • the searching for the collection attribute corresponding to the hash value to be searched based on the correspondence between the previously established hash value and the collection attribute may include:
  • the similarity between hash values there are many ways to calculate the similarity. For example, you can use the Hamming distance between hash values to calculate the similarity between hash values.
  • the hash values in the correspondence may be arranged according to the order of similarity from high to low, and then the preset number of hashes are selected as the hash value whose similarity satisfies the preset condition, and the selected hash value is selected.
  • the corresponding collection attribute is used as the collection attribute corresponding to the target feature to be searched.
  • the hash value with the largest similarity may be used as the hash value whose similarity satisfies the preset condition; or, the hash value whose similarity is greater than the preset threshold may be used as the hash whose similarity satisfies the preset condition. Value, etc., is not limited.
  • the method further includes: determining an acquisition attribute of the image to be identified as an acquisition attribute to be searched.
  • S103 includes: in a correspondence between the target feature and the collection attribute that is pre-established, searching for a target collection attribute that is smaller than a preset threshold value and a target feature corresponding to the target collection attribute, and Match target features;
  • the target collection attribute is used as an collection attribute corresponding to the target feature to be searched.
  • the search range is first narrowed by the collection attribute, and then further searched in the narrowed search range.
  • the movement trajectory of the human target to be identified is generally continuous, and the probability that the image having the same acquisition property contains the same human target is relatively large. Therefore, the application of the present embodiment is more accurate.
  • the target features or the face features stored in the corresponding relationship may be combined and processed periodically. It can be understood that if the similarity of the plurality of target/face features is high in the stored correspondence, it is considered that the plurality of target/face features belong to the same human target, and the plurality of target/face features can be combined into one Target feature group.
  • the target feature group matching the target feature to be searched is searched for.
  • the S103 includes: searching, in a correspondence relationship between the target feature and the collection attribute that is pre-established, a target feature group that matches the target feature to be searched; wherein the target feature group is each target belonging to the same human target And a collection attribute corresponding to each target feature included in the target feature group is used as an collection attribute corresponding to the target feature to be searched.
  • a facial feature set matching the facial features to be searched wherein the facial feature groups are each facial features belonging to the same human target
  • the collection attribute corresponding to each face feature included in the face feature group is used as an collection attribute corresponding to the face feature to be searched.
  • the correspondence between the target feature, the face feature, and the collection attribute is established.
  • the target feature and the face feature may be combined together to belong to the same human target.
  • the target feature and the face feature form a feature set.
  • a matching rule when it is determined whether the feature to be searched and the feature group match, a matching rule may be set. For example, the similarity between the feature to be searched and all features in the feature group is greater than a preset threshold, or the similarity between the feature to be searched and the feature in the feature group is greater than a preset threshold, and so on, and the specific matching rule is not Make a limit.
  • the embodiment of the present application further provides a human target trajectory determining device.
  • FIG. 4 is a schematic structural diagram of a human body target identification device according to an embodiment of the present disclosure, including: an obtaining module 401, configured to acquire an image to be processed; and a first extraction module 402, configured to extract, to be tracked, the image to be processed
  • the target feature of the human target is used as the target feature to be searched;
  • the first search module 403 is configured to search for the collection attribute corresponding to the target feature to be searched based on the corresponding relationship between the target feature and the collected attribute; wherein the corresponding The collection attribute corresponding to a target feature in the relationship is an acquisition attribute of the image having the target feature, the collection attribute includes a collection location, and the first determining module 404 is configured to determine the to-be-tracked according to the found collection attribute.
  • the trajectory of the human target is configured to acquire an image to be processed
  • a first extraction module 402 configured to extract, to be tracked, the image to be processed
  • the target feature of the human target is used as the target feature to be searched
  • the device may further include: a second extraction module and a second search module (not shown), wherein the second extraction module is configured to extract a target to be tracked in the image to be processed a face feature, as a face feature to be searched; a second search module, configured to search for an acquisition attribute corresponding to the face feature to be searched based on a correspondence between the pre-established face feature and the collection attribute;
  • the collection attribute corresponding to a face feature in the correspondence relationship is an acquisition attribute of an image having the face feature.
  • the device may further include: a third extraction module (not shown), configured to extract a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched;
  • the first searching module 403 is specifically configured to: search for an acquisition attribute that matches the to-be-find target feature and the to-be-find facial feature based on a pre-established correspondence between the target feature and the face feature, and the collection attribute.
  • an acquisition attribute corresponding to the target feature to be searched wherein, a corresponding target feature and a face feature belong to the same human target in the corresponding relationship; and a corresponding target feature and a face feature are collected in the corresponding relationship
  • the attribute is an acquisition attribute of an image having the target feature and the face feature.
  • the first extraction module 402 may be specifically configured to: extract an original target feature of the human target to be tracked in the image to be processed, and calculate a hash value of the original target feature as a hash value to be searched.
  • the first search module 403 is specifically configured to: search for the collection attribute corresponding to the hash value to be searched based on the correspondence between the pre-established hash value and the collection attribute.
  • the first searching module 403 may be specifically configured to: respectively calculate, between each hash value included in the correspondence between the pre-established hash value and the collection attribute, and the hash value to be searched. Similarity; determines an acquisition attribute corresponding to a hash value whose similarity satisfies a preset condition.
  • the device may further include: a second determining module (not shown), configured to determine an collection attribute of the to-be-processed image as an acquisition attribute to be searched; a first searching module 403, specifically The method may be configured to: search for a target collection attribute that is smaller than a preset threshold and a target feature corresponding to the target collection attribute, in a corresponding relationship between the target feature and the collection attribute, and the target feature corresponding to the target collection attribute. And determining whether the target feature to be matched matches the target feature to be searched; if yes, using the target acquisition attribute as the collection attribute corresponding to the target feature to be searched.
  • the first searching module 403 may be specifically configured to: in a correspondence between the target feature and the collection attribute that are pre-established, search for a target feature group that matches the target feature to be searched; wherein the target The feature set is composed of each target feature belonging to the same human target; the collection attribute corresponding to each target feature included in the target feature set is used as the collection attribute corresponding to the target feature to be searched.
  • the collection attribute may further include an acquisition moment.
  • the embodiment of the present application further provides an electronic device, as shown in FIG. 5, including a processor 501, a communication interface 502, a memory 503, and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 pass through the communication bus 504.
  • the memory 503 is configured to store the computer program
  • the processor 501 is configured to: when the program stored on the memory 503 is executed, the following steps are performed: acquiring the image to be processed; and extracting the human body to be tracked in the image to be processed A target feature of the target, as a target feature to be searched; searching for an acquisition attribute corresponding to the target feature to be searched based on a correspondence between the previously established target feature and the collected attribute; wherein, the corresponding feature in the corresponding relationship is collected
  • the attribute is an acquisition attribute of an image having the target feature, the collection attribute includes a collection location; and the trajectory of the human target to be tracked is determined according to the found collection attribute.
  • the processor 501 may be further configured to: after the step of acquiring an image to be processed, extract a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched Searching for the collection attribute corresponding to the face feature to be searched based on the correspondence between the pre-established face feature and the collection attribute; wherein the collection attribute corresponding to a face feature in the correspondence relationship has the face feature The collection properties of the image.
  • the processor 501 may be further configured to: extract a facial feature of the to-be-tracked human target in the image to be processed, as a facial feature to be searched; and the pre-established target feature and collection And the step of searching for the collection attribute corresponding to the target feature to be searched, including: searching and selecting the target feature to be searched based on the corresponding relationship between the target feature and the face feature and the collected attribute
  • the collection attribute of the face feature matching to be searched is used as the collection attribute corresponding to the target feature to be searched; wherein a corresponding target feature and the face feature belong to the same human target in the corresponding relationship;
  • the acquisition attribute corresponding to the target feature and the face feature is an acquisition attribute of the image having the target feature and the face feature.
  • the step of extracting a target feature of the human target to be tracked in the image to be processed, as a target feature to be searched includes: extracting an original target feature of the human target to be tracked in the image to be processed, and calculating a hash value of the original target feature as a hash value to be searched; the step of searching for an acquisition attribute corresponding to the target feature to be searched based on the correspondence between the pre-established target feature and the collected attribute, including: Corresponding relationship between the established hash value and the collected attribute, and searching for the collection attribute corresponding to the hash value to be searched.
  • the step of searching for the collection attribute corresponding to the hash value to be searched according to the correspondence between the pre-established hash value and the collection attribute includes: separately calculating a pre-established hash value and an acquisition attribute The similarity between each hash value included in the correspondence relationship and the hash value to be searched; and the collection attribute corresponding to the hash value whose similarity satisfies the preset condition is determined.
  • the processor 501 may be further configured to: after the step of acquiring an image to be processed, determining an collection attribute of the image to be processed as an attribute to be searched; the pre-established And the step of searching for the collection attribute corresponding to the target feature to be searched, the method comprising: searching for a difference between the target attribute and the collection attribute in a pre-established correspondence relationship between the target feature and the collection attribute a target acquisition attribute that is smaller than a preset threshold, and a target feature corresponding to the target collection attribute, as a target feature to be matched; determining whether the target feature to be matched matches the target feature to be searched; if yes, the target The collection attribute is used as an collection attribute corresponding to the target feature to be searched.
  • the step of searching for the collection attribute corresponding to the target feature to be searched according to the correspondence between the pre-established target feature and the collection attribute includes: in the correspondence between the pre-established target feature and the collection attribute And searching for a target feature group that matches the target feature to be searched; wherein the target feature group is composed of each target feature belonging to the same human target; and collecting corresponding target features included in the target feature set
  • the attribute is used as an collection attribute corresponding to the target feature to be searched.
  • the collection attribute may further include an acquisition moment.
  • the communication bus mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the communication bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used for communication between the above electronic device and other devices.
  • the memory may include a random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.
  • RAM random access memory
  • NVM non-volatile memory
  • the memory may also be at least one storage device located away from the aforementioned processor.
  • the above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; or may be a digital signal processing (DSP), dedicated integration.
  • CPU central processing unit
  • NP network processor
  • DSP digital signal processing
  • ASIC Application Specific Integrated Circuit
  • FPGA Field-Programmable Gate Array
  • the embodiment of the present application provides a computer readable storage medium, where the computer readable storage medium stores a computer program, and when the computer program is executed by the processor, the following steps are performed: acquiring an image to be processed; and extracting the image to be processed
  • the target feature of the human target to be tracked is used as the target feature to be searched; based on the correspondence between the pre-established target feature and the collected attribute, the collection attribute corresponding to the target feature to be searched is searched; wherein, a target in the corresponding relationship
  • the collection attribute corresponding to the feature is an acquisition attribute of the image having the target feature, the collection attribute includes a collection location, and the trajectory of the human target to be tracked is determined according to the found collection attribute.
  • the following steps may be further implemented: after the step of acquiring an image to be processed, extracting a facial feature of the human target to be tracked in the image to be processed, as a Searching for a face feature; searching for an acquisition attribute corresponding to the face feature to be searched based on a correspondence between the pre-established face feature and the collection attribute; wherein the collection attribute corresponding to a face feature in the correspondence relationship has The collection property of the image of the face feature.
  • the following steps may be further implemented: extracting a facial feature of the human target to be tracked in the image to be processed, as a facial feature to be searched; And the step of searching for the collection attribute corresponding to the target feature to be searched, including: searching for the target to be searched based on the corresponding relationship between the target feature and the face feature and the collected attribute And a collection attribute corresponding to the feature to be searched as the collection attribute corresponding to the target feature to be searched; wherein a corresponding target feature and a face feature in the correspondence relationship belong to the same human target;
  • the acquisition attribute corresponding to a target feature and the face feature in the correspondence relationship is an acquisition attribute of the image having the target feature and the face feature.
  • the step of extracting a target feature of the human target to be tracked in the image to be processed, as a target feature to be searched includes: extracting an original target feature of the human target to be tracked in the image to be processed, and calculating a hash value of the original target feature as a hash value to be searched; the step of searching for an acquisition attribute corresponding to the target feature to be searched based on the correspondence between the pre-established target feature and the collected attribute, including: Corresponding relationship between the established hash value and the collected attribute, and searching for the collection attribute corresponding to the hash value to be searched.
  • the step of searching for the collection attribute corresponding to the hash value to be searched according to the correspondence between the pre-established hash value and the collection attribute includes: separately calculating a pre-established hash value and an acquisition attribute The similarity between each hash value included in the correspondence relationship and the hash value to be searched; and the collection attribute corresponding to the hash value whose similarity satisfies the preset condition is determined.
  • the following steps may be further implemented: after the step of acquiring an image to be processed, determining an collection attribute of the image to be processed as an attribute to be searched; And the step of searching for the collection attribute corresponding to the target feature to be searched according to the pre-established correspondence between the target feature and the collection attribute, including: searching and collecting the to-be-searched in the correspondence between the pre-established target feature and the collection attribute And determining, by the target acquisition attribute that is smaller than the preset threshold, the target acquisition attribute corresponding to the target collection attribute, and the target feature corresponding to the target collection attribute, and determining whether the target feature to be matched matches the target feature to be searched; if yes, The target collection attribute is used as an collection attribute corresponding to the target feature to be searched.
  • the step of searching for the collection attribute corresponding to the target feature to be searched according to the correspondence between the pre-established target feature and the collection attribute includes: in the correspondence between the pre-established target feature and the collection attribute And searching for a target feature group that matches the target feature to be searched; wherein the target feature group is composed of each target feature belonging to the same human target; and collecting corresponding target features included in the target feature set
  • the attribute is used as an collection attribute corresponding to the target feature to be searched.
  • the collection attribute may further include an acquisition moment.
  • the embodiment of the present application also discloses an executable program code for being executed to implement any of the above-described human target trajectory determining methods.
  • the various embodiments in the present specification are described in a related manner, and the same or similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments.
  • the apparatus embodiment shown in FIG. 4, the electronic device embodiment shown in FIG. 5, the above computer readable storage medium embodiment, and the above executable program code embodiment are basically similar to FIG. 3 is a method embodiment, so the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment shown in FIG. 1-3.

Abstract

本申请实施例提供了一种人体目标轨迹确定方法及装置，方法包括：提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方案，提高了确定人员轨迹的准确性。

Description

一种人体目标轨迹确定方法及装置
本申请要求于2017年8月31日提交中国专利局、申请号为201710770254.5、发明名称为“一种人体目标轨迹确定方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图像处理技术领域，特别是涉及一种人体目标轨迹确定方法及装置。
背景技术
在日常生活中，有很多情况需要对人员轨迹进行追踪，比如，当发生盗窃、抢劫等事件时，需要对事件相关人员的轨迹进行追踪。
目前，通常利用人脸抓拍机追踪人员轨迹：在道路、公共场所等各种位置设置人脸抓拍机，通过人脸抓拍机采集到的人脸图像确定特定人员出现的时间、地点，由此对该特定人员的轨迹进行追踪。
但是，应用这种方案，在一些场景中，比如在天黑或者设备清晰度低的情况下，采集到的人脸图像不清晰，不能准确地识别人员身份，因而也就不能准确地确定人员轨迹。
发明内容
本申请实施例的目的在于提供一种人体目标轨迹确定方法及装置，以提高确定人员轨迹的准确性。
为达到上述目的，本申请实施例提供了一种人体目标轨迹确定方法，包括：获取待处理图像；提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
可选的，在所述获取待处理图像的步骤之后，还可以包括：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；基于预先建 立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
可选的，在所述获取待处理图像的步骤之后，还可以包括：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，可以包括：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
可选的，所述提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征的步骤，可以包括：提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，可以包括：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
可选的，所述基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性的步骤，可以包括：分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
可选的，在所述获取待处理图像的步骤之后，还可以包括：确定所述待处理图像的采集属性，作为待查找采集属性；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，可以包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；判断所述待匹配目标特征与所述待查找目标特征是否匹配；如果是，将所述目标采集属性作为所述待查找目标 特征对应的采集属性。
可选的，所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，可以包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
可选的，所述采集属性还包含采集时刻。
为达到上述目的，本申请实施例还提供了一种人体目标轨迹确定装置，包括：获取模块，用于获取待处理图像；第一提取模块，用于提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；第一查找模块，用于基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；第一确定模块，用于根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
可选的，所述装置还可以包括：第二提取模块，用于提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；第二查找模块，用于基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
可选的，所述装置还可以包括：第三提取模块，用于提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；所述第一查找模块，具体可以用于：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
可选的，所述第一提取模块，具体可以用于：提取所述待处理图像中待 追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；所述第一查找模块，具体可以用于：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
可选的，所述第一查找模块，具体可以用于：分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
可选的，所述装置还可以包括：第二确定模块，用于确定所述待处理图像的采集属性，作为待查找采集属性；所述第一查找模块，具体可以用于：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；判断所述待匹配目标特征与所述待查找目标特征是否匹配；如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
可选的，所述第一查找模块，具体可以用于：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
可选的，所述采集属性还包含采集时刻。
为达到上述目的，本申请实施例还提供了一种电子设备，包括处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；存储器，用于存放计算机程序；处理器，用于执行存储器上所存放的程序时，实现上述任一种人体目标轨迹确定方法。
为达到上述目的，本申请实施例还提供了一种计算机可读存储介质，所述计算机可读存储介质内存储有计算机程序，所述计算机程序被处理器执行时实现上述任一种人体目标轨迹确定方法。
为达到上述目的，本申请实施例还提供了一种可执行程序代码，所述可执行程序代码用于被运行以实现上述任一种人体目标轨迹确定方法。
应用本申请所示实施例，提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方案，提高了确定人员轨迹的准确性。
当然，实施本申请的任一产品或方法并不一定需要同时达到以上所述的所有优点。
附图说明
为了更清楚地说明本申请实施例和现有技术的技术方案，下面对实施例和现有技术中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的一种人体目标轨迹确定方法的流程示意图；
图2为本申请实施例提供的一种应用场景示意图；
图3为在图2应用场景中确定出的轨迹示意图；
图4为本申请实施例提供的一种人体目标轨迹确定装置的结构示意图；
图5为本申请实施例提供的一种电子设备的结构示意图。
具体实施方式
为使本申请的目的、技术方案、及优点更加清楚明白，以下参照附图并举实施例，对本申请进一步详细说明。显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
为了解决上述技术问题，本申请实施例提供了一种人体目标轨迹确定方法及装置。该方法及装置可以应用于具有图像处理功能的设备，具体不做限定。
下面首先对本申请实施例提供的一种人体目标轨迹确定方法进行详细说明。
图1为本申请实施例提供的一种人体目标轨迹确定的流程示意图，包括：
S101：获取待处理图像。
S102：提取该待处理图像中待追踪人体目标的目标特征，作为待查找目标特征。
S103：基于预先建立的目标特征与采集属性的对应关系，查找该待查找目标特征对应的采集属性；其中，该对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，该采集属性包含采集地点。
S104：根据所查找到的采集属性，确定该待追踪人体目标的轨迹。
应用本申请图1所示实施例，提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方案，提高了确定人员轨迹的准确性。
下面对图1所示实施进行详细说明：
S101：获取待处理图像。该待处理图像即为包含待追踪人体目标的图像。
作为一种实施方式，S101可以包括：接收用户输入的待处理图像；或者，作为另一种实施方式，S101可以包括：从指定采集设备中获取待处理图像。
可以理解，当需要追踪某人体目标的轨迹时，用户可以输入包含该人体目标的图像；或者，可以确定采集到该人体目标的采集设备，从该采集设备中获取包含该人体目标的图像。
或者，也可以采用其他方式获取待处理图像，具体不做限定。
S102：提取该待处理图像中待追踪人体目标的目标特征，作为待查找目标特征。
人体目标的目标特征可以包含颜色、纹理、尺寸等特征，也可以包含所穿戴的服饰特征，比如，是否背包、衣裤类型等，也可以包含人体目标的身高、体型等特征，具体不做限定。
在图像中提取目标特征的方式有很多，比如，利用边缘检测算法，检测图像中的人体目标区域，再提取该区域的图像特征；或者，利用预先训练得到的神经网络提取图像中人体目标的目标特征，等等，具体不做限定。
S103：基于预先建立的目标特征与采集属性的对应关系，查找该待查找目标特征对应的采集属性；其中，该对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，该采集属性包含采集地点。
可以采用如下步骤建立该对应关系：
获取采集设备采集的图像、以及图像对应的采集属性；
针对所获取的每张图像，提取该图像中每个人体目标的目标特征；
建立所述目标特征与所述采集属性的对应关系。
第一种情况，采集属性可以仅包含采集地点；第二种情况，采集属性可以包含采集地点及采集时刻。
对第一种情况进行举例说明：一台或多台采集设备可以与一台服务器通信连接，这些采集设备可以实时地将采集的图像发送给该服务器，服务器接收图像、提取图像中人体目标的目标特征，并根据接收时间的顺序，依次存储各张图像的目标特征。服务器可以确定接收到的每张图像的采集地点，并建立同一张图像的采集地点与目标特征的对应关系。
服务器确定接收到的每张图像的采集地点有多种方式，比如，预先获取各台设备的采集地点信息，并建立设备标识与采集地点的对应关系，当接收到某设备发送的图像时，确定该设备标识对应的采集地点，作为该图像的采集地点；或者，采集设备可以将采集到的图像与自身采集地点信息一并发送给服务器；等等，不再一一列举。
如图2所示的应用场景，该场景中的多台采集设备与同一服务器通信连接，服务器先接收到采集设备A发送的图像A1，提取A1中人体目标的目标特征A2， 确定图像A1的采集地点为A3，建立目标特征A2与采集地点A3的对应关系；然后，服务器又接收到采集设备B发送的图像B1，提取B1中人体目标的目标特征B2，并确定图像B1的采集地点为B3，建立目标特征B2与采集地点B3的对应关系；之后，服务器还接收到采集设备C发送的图像C1，提取C1中人体目标的目标特征C2，并确定图像C1的采集地点为C3，建立目标特征C2与采集地点C3的对应关系；则服务器建立的对应关系可以如表1所示：
表1
目标特征 采集地点
C2 C3
B2 B3
A2 A3
可见，表1中按照接收时间由晚至早的顺序，存储各对对应关系，或者，也可以按照接收时间由早至晚的顺序，存储各对对应关系，具体不做限定。
建立的对应关系(如表1)可以存储至服务器本地数据库中，或者存储至与服务器相连的数据库中。本申请实施例的执行主体与该服务器可以为同一设备，也可以为不同设备。
表1仅为一个简单的例子，并不对实际建立的对应关系构成限定。
对第二种情况进行举例说明：一台或多台采集设备可以与一台服务器通信连接，这些采集设备可以实时地、或者非实时地、将采集的图像发送给该服务器；服务器接收图像、提取图像中人体目标的目标特征，并确定接收到的每张图像的采集时刻采集地点，并建立同一张图像的采集时刻、采集地点与目标特征的对应关系。
如果采集设备实时地将采集到的图像发送给服务器，服务器可以将接收到图像的时刻作为采集时刻；或者，采集设备也可以将采集到的图像及该图像的采集时刻一并发送给服务器。
如果采集设备非实时地将采集到的图像发送给服务器，采集设备则将采集到的图像及该图像的采集时刻一并发送给服务器。
服务器确定接收到的每张图像的采集地点有多种方式，比如，预先获取各台设备的采集地点信息，并建立设备标识与采集地点的对应关系，当接收到某设备发送的图像时，确定该设备标识对应的采集地点，作为该图像的采集地点；或者，采集设备可以将采集到的图像与自身采集地点信息一并发送给服务器；等等，不再一一列举。
如图2所示的应用场景，该场景中的多台采集设备与同一服务器通信连接，服务器接收到采集设备A发送的图像A1，提取A1中人体目标的目标特征A2，确定图像A1的采集地点为A3、采集时刻为A4，建立目标特征A2与采集地点A3、采集时刻A4的对应关系；服务器接收到采集设备B发送的图像B1，提取B1中人体目标的目标特征B2，并确定图像B1的采集地点为B3、采集时刻为B4，建立目标特征B2与采集地点B3、采集时刻B4的对应关系；服务器接收到采集设备C发送的图像C1，提取C1中人体目标的目标特征C2，并确定图像C1的采集地点为C3、采集时刻为A4，建立目标特征C2与采集地点C3、采集时刻C4的对应关系；则服务器建立的对应关系可以如表2所示：
表2
目标特征 采集地点 采集时刻
C2 C3 C4
B2 B3 B4
A2 A3 A4
上述建立的对应关系(如表2)可以存储至服务器本地数据库中，或者存储至与服务器相连的数据库中。本申请实施例的执行主体与该服务器可以为同一设备，也可以为不同设备。
表2仅为一个简单的例子，并不对实际建立的对应关系构成限定。
S104：根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
在上述第二种情况中，采集属性包含采集时刻及采集地点，根据时刻及地点，可以确定出轨迹。以图3所示场景进行说明：假设在所建立的对应关系中查找到同一人体目标的三条采集属性，分别为：1、采集时刻：2017年7月14日上午9点5分、采集地点：B3，2、采集时刻：2017年7月14日上午9点3分、采集地点：C3，3、采集时刻：2017年7月14日上午9点、采集地点：F3，则可以确定出该人员的轨迹如图3中虚线所示，由F3至C3，再至B3。
而在上述第一种情况下，虽然采集属性中不包含采集时间，但由于对应关系按照接收图像的按照接收时间的顺序来存储，也可以根据采集地点及该顺序确定出轨迹。以图3所示场景进行说明：假设在所建立的对应关系中查找到同一人体目标的三条采集属性，依次为：1、采集地点：B3，2、采集地点：C3，3、采集地点：F3。假设对应关系的存储顺序为按照接收时间由晚至早的顺序来存储，也就是说，这三条采集属性中，接收时间最早的第三条，最晚的是第一条，因此可以确定出该人员的轨迹如图3中虚线所示，由F3至C3，再至B3。
作为一种实施方式，还可以建立人脸特征与采集属性的对应关系，这种情况下，在S101之后，还可以包括：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
具体的，建立人脸特征与采集属性的对应关系可以包括：
获取采集设备采集的图像、以及图像对应的采集属性；
针对所获取的每张图像，提取该图像中每个人体目标的人脸特征；
建立所述人脸特征与所述采集属性的对应关系。
建立人脸特征与采集属性的对应关系、与建立目标特征与采集属性的对应关系的过程类似，不再赘述。
本实施方式中，待处理图像中不仅包含人体目标的目标特征，还包含人体目标的人脸特征；而且除上述目标特征与采集属性的对应关系外，还建立了人脸特征与采集属性的对应关系；这样，在图1所示实施例的基础上，还可以根据人脸特征，查找到更多采集属性，基于更多采集属性，确定出更准确的轨迹。
作为一种实施方式，可以建立目标特征、人脸特征、采集属性三者的对应关系，具体过程可以包括：
获取采集设备采集的图像、以及图像对应的采集属性；
针对所获取的每张图像，提取该图像中每个人体目标的目标特征及人脸特征；
建立所述目标特征及人脸特征与所述采集属性的对应关系；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标。
建立三者的对应关系、与建立目标特征与采集属性的对应关系的过程类似，不再赘述。
这种情况下，在S101之后，还可以包括：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
S103可以包括：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性。
查找与待查找目标特征及待查找人脸特征匹配的采集属性时，可以设定匹配规则。比如，对应关系中的目标特征与待查找目标特征的相似度大于第一预设阈值，并且对应关系中的人脸特征与及待查找人脸特征的相似度大于第二预设阈值，第一预设阈值与第二预设阈值可以根据实际情况设定，可以相同或不同。具体的匹配规则不做限定。
需要说明的是，如果待处理图像中包含多个待追踪人体目标，则可以针对每个待追踪人体目标，提取其目标特征及人脸特征，作为待查找目标特征 及待查找人脸特征，在所建立的三者的对应关系中，查找与该待查找目标特征及待查找人脸特征匹配的采集属性。也就是说，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性时，所述待查找目标特征及所述待查找人脸特征属于同一人体目标，才是合理的。
应用本实施方式，查找与目标特征、人脸特征都匹配的采集属性，可以提高查找的准确性。
作为一种实施方式，人体目标的目标特征、以及人脸特征可以用哈希值来表示。这种实施方式中，S102中提取的目标特征为哈希值，预先建立的对应关系中的目标特征也为哈希值。
具体的，S103可以包括：提取所述待处理图像中待识别人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值。
提取人脸特征可以包括：提取所述待处理图像中待识别人体目标的原始人脸特征，计算所述原始人脸特征的哈希值，作为待查找哈希值。
S104或者查找人脸特征对应的采集属性可以包括：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
应用本实施方式，用哈希值来表示目标特征、人脸特征，可以提高查找效率。
在本实施方式中，基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性，可以包括：
分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
计算相似度的方式有很多，比如可以利用哈希值之间的汉明距离，计算哈希值之间的相似度。可以按照相似度由高到低的顺序，排列该对应关系中的各哈希值，然后选取前预设数量个哈希值作为相似度满足预设条件的哈希值，将选取的哈希值对应的采集属性作为待查找目标特征对应的采集属性。
或者，也可以仅将相似度最大的哈希值作为相似度满足预设条件的哈希值；或者，也可以将相似度大于预设阈值的哈希值作为相似度满足预设条件的哈希值，等等，具体不做限定。
作为一种实施方式，在S101之后，还可以包括：确定所述待识别图像的采集属性，作为待查找采集属性。
S103包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；
判断所述待匹配目标特征与所述待查找目标特征是否匹配；
如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
在本实施方式中，先利用采集属性缩小查找范围，然后在缩小后的查找范围中作进一步的查找。
可以理解，如果采用计算哈希值相似度的查找方式，本实施方式中，不需要计算待查找哈希值与对应关系中所有哈希值的相似度，而是先根据采集属性过滤掉一部分哈希值，仅计算待查找哈希值与剩余部分哈希值的相似度，降低了计算量，进一步提高了查找效率。
再者，待识别人体目标的移动轨迹一般是连续的，采集属性较相近的图像中包含同一人体目标的概率较大，因此，应用本实施方式查找更准确。
作为一种实施方式，在建立上述几种对应关系后，可以周期性地将对应关系中存储的目标特征或者人脸特征进行合并处理。可以理解，如果存储的对应关系中，多份目标/人脸特征的相似度较高，则认为这多份目标/人脸特征属于同一人体目标，可以将这多份目标/人脸特征组成一个目标特征组。
这种情况下，在对应关系中查找待查找目标特征，则为查找与待查找目标特征匹配的目标特征组。具体的，S103包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中， 所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性，包括：
在预先建立的人脸特征与采集属性的对应关系中，查找与所述待查找人脸特征匹配的人脸特征组；其中，所述人脸特征组由属于同一人体目标的各份人脸特征组成；将所述人脸特征组中包含的各份人脸特征对应的采集属性作为所述待查找人脸特征对应的采集属性。
在上述一种实施方式中，建立了目标特征、人脸特征、以及采集属性三者的对应关系，这种实施方式中，可以将目标特征、人脸特征一起进行合并，将属于同一人体目标的目标特征、人脸特征组成一个特征组。
基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，包括：
在预先建立的目标特征与人脸特征、以及采集属性的对应关系中，查找与所述待查找人脸特征及所述待查找人脸特征匹配的特征组；其中，所述特征组由属于同一人体目标的各份目标特征及人脸特征组成；将所述人脸特征组中包含的各份目标特征及人脸特征对应的采集属性作为所述待查找人脸特征对应的采集属性。
本实施方式中，判断待查找特征与特征组是否匹配时，可以设定匹配规则。比如，待查找特征与特征组中所有特征的相似度均大于预设阈值，或者，待查找特征与特征组中特征的相似度大于预设阈值的数量较多，等等，具体的匹配规则不做限定。
应用本申请图1所示实施例，提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方 案，提高了确定人员轨迹的准确性。
与上述方法实施例相对应，本申请实施例还提供一种人体目标轨迹确定装置。
图4为本申请实施例提供的一种人体目标身份识别装置的结构示意图，包括：获取模块401，用于获取待处理图像；第一提取模块402，用于提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；第一查找模块403，用于基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；第一确定模块404，用于根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
作为一种实施方式，所述装置还可以包括：第二提取模块和第二查找模块(图中未示出)，其中，第二提取模块，用于提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；第二查找模块，用于基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
作为一种实施方式，所述装置还可以包括：第三提取模块(图中未示出)，用于提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；第一查找模块403，具体可以用于：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
作为一种实施方式，第一提取模块402，具体可以用于：提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值， 作为待查找哈希值；第一查找模块403，具体可以用于：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
作为一种实施方式，第一查找模块403，具体可以用于：分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
作为一种实施方式，所述装置还可以包括：第二确定模块(图中未示出)，用于确定所述待处理图像的采集属性，作为待查找采集属性；第一查找模块403，具体可以用于：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；判断所述待匹配目标特征与所述待查找目标特征是否匹配；如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，第一查找模块403，具体可以用于：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，所述采集属性还可以包含采集时刻。
应用本申请图4所示实施例，提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方案，提高了确定人员轨迹的准确性。
本申请实施例还提供了一种电子设备，如图5所示，包括处理器501、通信接口502、存储器503和通信总线504，其中，处理器501，通信接口502，存储器503通过通信总线504完成相互间的通信，存储器503，用于存放计算机程 序；处理器501，用于执行存储器503上所存放的程序时，实现如下步骤：获取待处理图像；提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
作为一种实施方式，处理器501还可以用于实现如下步骤：在所述获取待处理图像的步骤之后，提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
作为一种实施方式，处理器501还可以用于实现如下步骤：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
作为一种实施方式，所述提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征的步骤，包括：提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
作为一种实施方式，所述基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性的步骤，包括：分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间 的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
作为一种实施方式，处理器501还可以用于实现如下步骤：在所述获取待处理图像的步骤之后，确定所述待处理图像的采集属性，作为待查找采集属性；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；判断所述待匹配目标特征与所述待查找目标特征是否匹配；如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，所述采集属性还可以包含采集时刻。
上述电子设备提到的通信总线可以是外设部件互连标准(Peripheral Component Interconnect，PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture，EISA)总线等。该通信总线可以分为地址总线、数据总线、控制总线等。为便于表示，图中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
通信接口用于上述电子设备与其他设备之间的通信。
存储器可以包括随机存取存储器(Random Access Memory，RAM)，也可以包括非易失性存储器(Non-Volatile Memory，NVM)，例如至少一个磁盘存储器。可选的，存储器还可以是至少一个位于远离前述处理器的存储装置。
上述的处理器可以是通用处理器，包括中央处理器(Central Processing Unit，CPU)、网络处理器(Network Processor，NP)等；还可以是数字信号处理器(Digital Signal Processing，DSP)、专用集成电路(Application Specific  Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
应用本申请图5所示实施例，提取图像中待追踪人体目标的目标特征，作为待查找目标特征，基于预先建立的目标特征与采集属性的对应关系，查找待查找目标特征对应的采集属性，根据所查找到的采集属性，确定待追踪人员的轨迹；可见，本方案中，不需要利用人脸图像确定人员轨迹，即使采集到的人脸图像不清晰，也不会降低确定人员轨迹的准确性；因此，应用本方案，提高了确定人员轨迹的准确性。
本申请实施例提供一种计算机可读存储介质，所述计算机可读存储介质内存储有计算机程序，所述计算机程序被处理器执行时实现如下步骤：获取待处理图像；提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
作为一种实施方式，所述计算机程序被处理器执行时还可以实现如下步骤：在所述获取待处理图像的步骤之后，提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
作为一种实施方式，所述计算机程序被处理器执行时还可以实现如下步骤：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对 应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
作为一种实施方式，所述提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征的步骤，包括：提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
作为一种实施方式，所述基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性的步骤，包括：分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；确定相似度满足预设条件的哈希值对应的采集属性。
作为一种实施方式，所述计算机程序被处理器执行时还可以实现如下步骤：在所述获取待处理图像的步骤之后，确定所述待处理图像的采集属性，作为待查找采集属性；所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；判断所述待匹配目标特征与所述待查找目标特征是否匹配；如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
作为一种实施方式，所述采集属性还可以包含采集时刻。
本申请实施例还公开了一种可执行程序代码，所述可执行程序代码用于 被运行以实现上述任一种人体目标轨迹确定方法。
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
本说明书中的各个实施例均采用相关的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于图4所示的装置实施例、图5所示的电子设备实施例、上述计算机可读存储介质实施例、以及上述可执行程序代码实施例而言，由于其基本相似于图1-3所示的方法实施例，所以描述的比较简单，相关之处参见图1-3所示的方法实施例的部分说明即可。
以上所述仅为本申请的较佳实施例而已，并非用于限定本申请的保护范围。凡在本申请的精神和原则之内所作的任何修改、等同替换、改进等，均包含在本申请的保护范围内。

Claims (24)

  1. 一种人体目标轨迹确定方法，其特征在于，包括：
    获取待处理图像；
    提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；
    基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；
    根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
  2. 根据权利要求1所述的方法，其特征在于，在所述获取待处理图像的步骤之后，还包括：
    提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
    基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
  3. 根据权利要求1所述的方法，其特征在于，在所述获取待处理图像的步骤之后，还包括：
    提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
  4. 根据权利要求1所述的方法，其特征在于，所述提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征的步骤，包括：
    提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
  5. 根据权利要求4所述的方法，其特征在于，所述基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性的步骤，包括：
    分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；
    确定相似度满足预设条件的哈希值对应的采集属性。
  6. 根据权利要求1所述的方法，其特征在于，在所述获取待处理图像的步骤之后，还包括：
    确定所述待处理图像的采集属性，作为待查找采集属性；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；
    判断所述待匹配目标特征与所述待查找目标特征是否匹配；
    如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
  7. 根据权利要求1所述的方法，其特征在于，所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步 骤，包括：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；
    将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
  8. 根据权利要求1-7任一项所述的方法，其特征在于，所述采集属性还包含采集时刻。
  9. 一种人体目标轨迹确定装置，其特征在于，包括：
    获取模块，用于获取待处理图像；
    第一提取模块，用于提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；
    第一查找模块，用于基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；
    第一确定模块，用于根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
  10. 根据权利要求9所述的装置，其特征在于，所述装置还包括：
    第二提取模块，用于提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
    第二查找模块，用于基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
  11. 根据权利要求9所述的装置，其特征在于，所述装置还包括：
    第三提取模块，用于提取所述待处理图像中待追踪人体目标的人脸特征， 作为待查找人脸特征；
    所述第一查找模块，具体用于：
    基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
  12. 根据权利要求9所述的装置，其特征在于，所述第一提取模块，具体用于：
    提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；
    所述第一查找模块，具体用于：
    基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
  13. 根据权利要求12所述的装置，其特征在于，所述第一查找模块，具体用于：
    分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；
    确定相似度满足预设条件的哈希值对应的采集属性。
  14. 根据权利要求9所述的装置，其特征在于，所述装置还包括：
    第二确定模块，用于确定所述待处理图像的采集属性，作为待查找采集属性；
    所述第一查找模块，具体用于：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；
    判断所述待匹配目标特征与所述待查找目标特征是否匹配；
    如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
  15. 根据权利要求9所述的装置，其特征在于，所述第一查找模块，具体用于：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；
    将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
  16. 根据权利要求9-15任一项所述的装置，其特征在于，所述采集属性还包含采集时刻。
  17. 一种电子设备，其特征在于，包括处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；
    存储器，用于存放计算机程序；
    处理器，用于执行存储器上所存放的程序时，实现如下步骤：
    获取待处理图像；
    提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征；
    基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性；其中，所述对应关系中一份目标特征对应的采集属性为具有该目标特征的图像的采集属性，所述采集属性包含采集地点；
    根据所查找到的采集属性，确定所述待追踪人体目标的轨迹。
  18. 根据权利要求17所述的设备，其特征在于，所述处理器还用于实现如下步骤：
    在所述获取待处理图像的步骤之后，提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
    基于预先建立的人脸特征与采集属性的对应关系，查找所述待查找人脸特征对应的采集属性；其中，所述对应关系中一份人脸特征对应的采集属性为具有该人脸特征的图像的采集属性。
  19. 根据权利要求17所述的设备，其特征在于，所述处理器还用于实现如下步骤：提取所述待处理图像中待追踪人体目标的人脸特征，作为待查找人脸特征；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    基于预先建立的目标特征与人脸特征、以及采集属性的对应关系，查找与所述待查找目标特征及所述待查找人脸特征匹配的采集属性，作为所述待查找目标特征对应的采集属性；其中，所述对应关系中一份对应的目标特征与人脸特征属于同一人体目标；所述对应关系中一份目标特征与人脸特征对应的采集属性为具有该目标特征与人脸特征的图像的采集属性。
  20. 根据权利要求17所述的设备，其特征在于，所述提取所述待处理图像中待追踪人体目标的目标特征，作为待查找目标特征的步骤，包括：
    提取所述待处理图像中待追踪人体目标的原始目标特征，计算所述原始目标特征的哈希值，作为待查找哈希值；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性。
  21. 根据权利要求20所述的设备，其特征在于，所述基于预先建立的哈希值与采集属性的对应关系，查找所述待查找哈希值对应的采集属性的步骤，包括：
    分别计算预先建立的哈希值与采集属性的对应关系中所包括的各哈希值与所述待查找哈希值之间的相似度；
    确定相似度满足预设条件的哈希值对应的采集属性。
  22. 根据权利要求17所述的设备，其特征在于，所述处理器还用于实现如下步骤：在所述获取待处理图像的步骤之后，确定所述待处理图像的采集属性，作为待查找采集属性；
    所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找采集属性的差值小于预设阈值的目标采集属性、以及所述目标采集属性对应的目标特征，作为待匹配目标特征；
    判断所述待匹配目标特征与所述待查找目标特征是否匹配；
    如果是，将所述目标采集属性作为所述待查找目标特征对应的采集属性。
  23. 根据权利要求17所述的设备，其特征在于，所述基于预先建立的目标特征与采集属性的对应关系，查找所述待查找目标特征对应的采集属性的步骤，包括：
    在预先建立的目标特征与采集属性的对应关系中，查找与所述待查找目标特征匹配的目标特征组；其中，所述目标特征组由属于同一人体目标的各份目标特征组成；
    将所述目标特征组中包含的各份目标特征对应的采集属性作为所述待查找目标特征对应的采集属性。
  24. 根据权利要求17-23任一项所述的设备，其特征在于，所述采集属性还包含采集时刻。
PCT/CN2018/101666 2017-08-31 2018-08-22 一种人体目标轨迹确定方法及装置 WO2019042196A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US16/643,373 US11157720B2 (en) 2017-08-31 2018-08-22 Method and device for determining path of human target
EP18852621.4A EP3678057B1 (en) 2017-08-31 2018-08-22 Method and device for determining path of human target

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710770254.5A CN109426787A (zh) 2017-08-31 2017-08-31 一种人体目标轨迹确定方法及装置
CN201710770254.5 2017-08-31

Publications (1)

Publication Number Publication Date
WO2019042196A1 true WO2019042196A1 (zh) 2019-03-07

Family

ID=65504644

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/101666 WO2019042196A1 (zh) 2017-08-31 2018-08-22 一种人体目标轨迹确定方法及装置

Country Status (4)

Country Link
US (1) US11157720B2 (zh)
EP (1) EP3678057B1 (zh)
CN (1) CN109426787A (zh)
WO (1) WO2019042196A1 (zh)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SG10201807675TA (en) * 2018-09-06 2020-04-29 Nec Asia Pacific Pte Ltd Duration and Potential Region of Interest for Suspicious Activities
CN109376256B (zh) * 2018-09-29 2021-03-26 京东方科技集团股份有限公司 图像搜索方法及装置
WO2020136794A1 (ja) * 2018-12-27 2020-07-02 日本電気株式会社 情報処理システム、情報処理装置、情報処理方法、およびプログラム
CN110418076A (zh) * 2019-08-02 2019-11-05 新华智云科技有限公司 视频集锦生成方法、装置、电子设备及存储介质
CN110942003A (zh) * 2019-11-20 2020-03-31 中国建设银行股份有限公司 人员轨迹搜索方法及系统
US11361189B2 (en) * 2019-12-03 2022-06-14 Ping An Technology (Shenzhen) Co., Ltd. Image generation method and computing device
CN112541384B (zh) * 2020-07-30 2023-04-28 深圳市商汤科技有限公司 可疑对象查找方法及装置、电子设备及存储介质
CN112033454B (zh) * 2020-08-28 2022-07-12 杭州海康威视数字技术股份有限公司 轨迹数据处理方法和装置
CN111898592B (zh) * 2020-09-29 2020-12-29 腾讯科技（深圳）有限公司 一种轨迹数据处理方法、装置以及计算机可读存储介质
CN113158953B (zh) * 2021-04-30 2022-11-25 青岛海信智慧生活科技股份有限公司 一种人员查找方法、装置、设备及介质
KR102465437B1 (ko) * 2022-05-30 2022-11-10 (주)이앤제너텍 인공지능 기반 객체 추적 장치 및 방법

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104881637A (zh) * 2015-05-09 2015-09-02 广东顺德中山大学卡内基梅隆大学国际联合研究院 基于传感信息及目标追踪的多模信息系统及其融合方法
CN106599866A (zh) * 2016-12-22 2017-04-26 上海百芝龙网络科技有限公司 一种多维度用户身份识别方法
CN106874347A (zh) * 2016-12-26 2017-06-20 深圳市深网视界科技有限公司 一种匹配人体特征与mac地址的方法及系统

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2418312A (en) 2004-09-18 2006-03-22 Hewlett Packard Development Co Wide area tracking system
CN102687171B (zh) * 2009-10-16 2015-07-08 日本电气株式会社 人物检索装置及人物检索方法
US20110317010A1 (en) 2010-06-24 2011-12-29 Infosys Technologies Limited System and method for tracking a person in a pre-defined area
CN104794458A (zh) * 2015-05-07 2015-07-22 北京丰华联合科技有限公司 一种从模糊视频中识别人员的方法
CN105447466B (zh) * 2015-12-01 2019-07-23 深圳市图灵机器人有限公司 一种基于Kinect传感器的身份综合识别方法
CN105808709B (zh) * 2016-03-04 2019-10-29 智慧眼科技股份有限公司 人脸识别快速检索方法及装置
CN106650652A (zh) * 2016-12-14 2017-05-10 黄先开 基于人脸识别技术的轨迹跟踪系统及其方法
CN107016374A (zh) * 2017-04-12 2017-08-04 电子科技大学 面向特定目标的智能检测追踪及时空轨迹的生成方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104881637A (zh) * 2015-05-09 2015-09-02 广东顺德中山大学卡内基梅隆大学国际联合研究院 基于传感信息及目标追踪的多模信息系统及其融合方法
CN106599866A (zh) * 2016-12-22 2017-04-26 上海百芝龙网络科技有限公司 一种多维度用户身份识别方法
CN106874347A (zh) * 2016-12-26 2017-06-20 深圳市深网视界科技有限公司 一种匹配人体特征与mac地址的方法及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3678057A4

Also Published As

Publication number Publication date
EP3678057B1 (en) 2023-06-28
EP3678057A4 (en) 2020-07-29
EP3678057A1 (en) 2020-07-08
CN109426787A (zh) 2019-03-05
US11157720B2 (en) 2021-10-26
US20200257890A1 (en) 2020-08-13

Similar Documents

Publication Publication Date Title
WO2019042196A1 (zh) 一种人体目标轨迹确定方法及装置
WO2019042195A1 (zh) 一种人体目标身份识别方法及装置
US10068018B2 (en) System and method for determining whether a product image includes a logo pattern
WO2019085064A1 (zh) 医疗理赔拒付方法、装置、终端设备及存储介质
CN108038176B (zh) 一种路人库的建立方法、装置、电子设备及介质
WO2018099032A1 (zh) 一种目标跟踪方法及装置
CN105335726B (zh) 人脸识别置信度获取方法和系统
JP2018505495A (ja) 指紋重複領域の面積算出方法、それを行う電子機器、コンピュータプログラム、及び、記録媒体
WO2019001429A1 (zh) 一种多源数据融合方法和装置
WO2020082258A1 (zh) 一种多目标实时跟踪方法、装置及电子设备
CN111126122B (zh) 人脸识别算法评估方法及装置
CN103399896A (zh) 识别用户间关联关系的方法及系统
WO2019242672A1 (zh) 一种目标跟踪方法、装置及系统
WO2021238664A1 (zh) 信息采集方法和装置、关注度检测方法、装置和系统
CN103281341A (zh) 网络事件处理方法及装置
CN104699755A (zh) 一种基于数据挖掘的智能化多目标综合识别方法
CN110765760B (zh) 一种法律案件分配方法、装置、存储介质和服务器
US20200005075A1 (en) Computer Vision Systems and Methods for Automatically Detecting, Classifying, and Pricing Objects Captured in Images or Videos
JP2018512655A (ja) 製品インデキシング方法およびそのシステム
US20180276696A1 (en) Association method, and non-transitory computer-readable storage medium
WO2020073697A1 (zh) 图像识别的方法、装置、系统及计算设备
CN111368867A (zh) 档案归类方法及系统、计算机可读存储介质
Bostanci et al. A fuzzy brute force matching method for binary image features
CN111666441A (zh) 确定人员身份类型的方法、装置和电子系统
De Marsico et al. ES-RU: an e ntropy based rule to s elect r epresentative templates in face su rveillance

Legal Events

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

Ref document number: 18852621

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2018852621

Country of ref document: EP

Effective date: 20200331