WO2022193516A1 - Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur - Google Patents

Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur Download PDF

Info

Publication number
WO2022193516A1
WO2022193516A1 PCT/CN2021/107938 CN2021107938W WO2022193516A1 WO 2022193516 A1 WO2022193516 A1 WO 2022193516A1 CN 2021107938 W CN2021107938 W CN 2021107938W WO 2022193516 A1 WO2022193516 A1 WO 2022193516A1
Authority
WO
WIPO (PCT)
Prior art keywords
pedestrian
human body
image
area
images
Prior art date
Application number
PCT/CN2021/107938
Other languages
English (en)
Chinese (zh)
Inventor
王卫芳
龚国基
朱毅博
胡正
Original Assignee
奥比中光科技集团股份有限公司
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 奥比中光科技集团股份有限公司 filed Critical 奥比中光科技集团股份有限公司
Publication of WO2022193516A1 publication Critical patent/WO2022193516A1/fr

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
    • G06T7/246—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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • 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
    • G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
    • 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
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30232—Surveillance
    • 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 belongs to the field of machine vision, and in particular relates to a method and device for analyzing human flow based on a depth camera.
  • the main method for people flow analysis is to obtain the pedestrian image set in the area that needs to be analyzed by using a three-primary camera, and to achieve the purpose of pedestrian detection by identifying a single human feature (such as a face or head) contained in the pedestrian image set.
  • follow-up flow analysis is performed based on the detected pedestrian information.
  • obtaining color images of pedestrians through three-color cameras may violate the privacy of pedestrians, and color images often only reflect information in the plane dimension and cannot accurately track pedestrians.
  • the embodiments of the present application provide a method and device for analyzing pedestrian flow based on a depth camera, which can obtain a pedestrian image set through a depth camera, and import multiple frames of pedestrian images in the pedestrian image set into a human detection network, so as to determine the target of the depth camera.
  • the motion tracking data of pedestrian objects in the detection area is further generated to generate pedestrian flow analysis information.
  • the human detection network is obtained through deep learning training and can accurately track pedestrians, and the input of the human detection network is obtained by the depth camera.
  • the pedestrian image that is, the depth image, avoids the problem of the color image infringing the pedestrian's privacy, and solves the problem that the color image in the prior art can only reflect the information in the plane dimension and cannot accurately track the pedestrian.
  • an embodiment of the present application provides a method for analyzing pedestrian flow based on a depth camera, including: acquiring a pedestrian image set in a target detection area through a depth camera; the pedestrian image set includes multiple frames of pedestrian images; Import a human detection network, and output the human detection image corresponding to the pedestrian image in each frame; the human detection network is obtained through deep learning training; the target is determined based on the human detection images corresponding to the pedestrian images in multiple consecutive frames Motion tracking data of several pedestrian objects contained in the detection area; based on the motion tracking data of the several pedestrian objects, the analysis information of people flow about the target detection area is generated.
  • an embodiment of the present application provides a depth camera-based pedestrian flow analysis device, including: a pedestrian image acquisition module, configured to acquire a pedestrian image set related to a target detection area through the depth camera; the pedestrian image set includes multiple frame pedestrian images; a human body detection network module is used to import the pedestrian images into a human body detection network, and output the human body detection images corresponding to the pedestrian images in each frame; the human body detection network is obtained through deep learning training; motion tracking data a determination module, configured to determine motion tracking data of several pedestrian objects contained in the target detection area based on the human body detection images corresponding to the multiple frames of continuous pedestrian images; The motion tracking data of several pedestrian objects generates human flow analysis information about the target detection area.
  • an embodiment of the present application provides a terminal device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program When the method described in any one of the above-mentioned first aspect is implemented.
  • an embodiment of the present application provides a computer-readable storage medium, including: the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any one of the foregoing first aspects is implemented the method described.
  • an embodiment of the present application provides a computer program product that, when the computer program product runs on a terminal device, causes the terminal device to execute the method described in any one of the above-mentioned first aspects.
  • the pedestrian flow analysis method can obtain a pedestrian image set through a depth camera, and import multiple frames of pedestrian images in the pedestrian image set into a human detection network, so as to determine the target detection area of the depth camera.
  • the motion tracking data of pedestrian objects is used to further generate pedestrian flow analysis information.
  • the human detection network is obtained through deep learning training and can accurately track pedestrians.
  • the input of the human detection network is the pedestrian image obtained by the depth camera. That is, the depth image not only avoids the problem of easily infringing the privacy of pedestrians due to the use of color images in the analysis of the human flow, but also because the depth image can reflect the depth information, it can solve the problem of using color images to analyze the human flow in the prior art. It reflects the information in the plane dimension and cannot accurately track pedestrians.
  • Fig. 1 is the realization flow chart of the people flow analysis method that the first embodiment of the present application provides;
  • FIG. 2 is a schematic diagram of an application scenario provided by an embodiment of the present application.
  • Fig. 3 is the realization flow chart of the people flow analysis method provided by the second embodiment of the present application.
  • FIG. 4 is a schematic diagram of a human body detection network provided by an embodiment of the present application.
  • Fig. 5 is the realization flow chart of the human flow analysis method provided by the third embodiment of the present application.
  • Fig. 6 is the realization flow chart of the human flow analysis method provided by the fourth embodiment of the present application.
  • Fig. 7 is the realization flow chart of the human flow analysis method provided by the fifth embodiment of the present application.
  • Fig. 8 is the realization flow chart of the human flow analysis method provided by the sixth embodiment of the present application.
  • FIG. 9 is a schematic diagram of a monitoring screen effect provided by the sixth embodiment of the present application.
  • FIG. 10 is a schematic flowchart of a flow analysis method for people flow provided by an embodiment of the present application.
  • FIG. 11 is a schematic structural diagram of a human flow analysis device provided by an embodiment of the present application.
  • FIG. 12 is a schematic structural diagram of a terminal device provided by an embodiment of the present application.
  • the execution subject of the process is a terminal device.
  • the terminal device includes but is not limited to: a server, a computer, a smart phone, a tablet computer, and other devices capable of executing the people flow analysis method provided in this application.
  • Fig. 1 shows the realization flow chart of the people flow analysis method that the first embodiment of the present application provides, and details are as follows:
  • a pedestrian image set of the target detection area is acquired through a depth camera.
  • the pedestrian image set includes multiple frames of pedestrian images.
  • the depth camera is set at a position within the target detection area that can be observed.
  • the depth camera can specifically shoot down the target detection area where people flow statistics need to be performed, so that it can be viewed from above.
  • the moving trajectories of pedestrians appearing in the target detection area are obtained from the angle to conduct traffic statistics and analysis.
  • the shooting picture of the depth camera is specifically a picture about the target detection area. In order to subsequently detect pedestrian objects in the target detection area and generate pedestrian flow analysis information based on the detection results, multiple consecutive images of the target detection area are required, so the pedestrian image set is obtained through the depth camera.
  • the above method of acquiring the pedestrian image set may specifically be: controlling the depth camera to acquire multiple frames of pedestrian images corresponding to the target detection area at preset acquisition time intervals, and according to the acquisition time of each frame of pedestrian objects
  • the above sequence of pedestrian images is encapsulated by a preset number of consecutive pedestrian images to obtain the above pedestrian image set. For example, the pedestrian images collected within the same minute are encapsulated into a pedestrian image set.
  • the pedestrian image in the pedestrian image set is a depth image, and the pedestrian image only contains the depth value corresponding to each pixel, and does not contain the pixel value of the three primary colors of each pixel, so as to reduce the information of the input image of the subsequent human detection network. It can improve the efficiency of subsequent generation of people flow analysis information.
  • the pedestrian image is imported into a human body detection network, and a human body detection image corresponding to each frame of the pedestrian image is output.
  • the human body detection network is obtained through deep learning training; specifically, multiple frames of training images are collected, and each frame of training images is marked with a true-value human body region; the training image is imported into the human body detection network, based on The internal parameters of the human body detection network divide the predicted human body region in the training image, calculate the mask loss value of this training based on the predicted human body region and the true human body region, and adjust the human body detection network based on the mask loss value.
  • the mask loss value is less than the preset adjustment threshold, it is recognized that the above-mentioned human body detection network has been adjusted.
  • the human body detection image is recognized through the trained human body detection network.
  • the input of the human body detection network is the above-mentioned pedestrian image
  • the output is the above-mentioned human body detection image
  • the human body detection image is an image in which several human body regions are marked in the pedestrian image.
  • the human body detection image may include at least one human body area, that is, the human body detection network detects that there is a pedestrian object in the target detection area; the human body detection image may not include a human body area, that is, the human body detection network detects There is no pedestrian object in the target detection area.
  • the above-mentioned method of obtaining a human body detection image through a human body detection network may specifically include: the human body detection network includes a feature extraction layer, a region division layer, and a region identification layer; import the pedestrian image into the feature extraction layer.
  • the feature image obtains the feature image; import the feature image into the region division layer to obtain the mask area of each object in the pedestrian image; import the feature image into the region recognition layer to obtain the classification information of each object in the pedestrian image (in order to improve the The output efficiency of the human body detection network, the value range of the classification information only includes human body or non-human body); if the classification information of the object is human body, the mask area of the object is identified as the above-mentioned human body area; all human body areas are identified Mark the pedestrian image to obtain the above-mentioned human detection image. Exemplarily, all the human body regions are marked in the pedestrian image to obtain the above-mentioned human body detection image.
  • the depth value of each pixel in the pedestrian image except the pixels in the above-mentioned human body region is marked as null
  • the human body detection image is obtained, that is, in the human body detection image, only the pixels in the above-mentioned human body region are associated with corresponding depth values, then when the motion tracking data is determined according to the human body detection image, interference items can be reduced, and only the above-mentioned human body region is considered. ,Improve efficiency.
  • the above-mentioned feature extraction layer can be a convolutional neural network (Convolutional Neural Networks, CNN), and the above-mentioned region division layer and region identification layer can be a part of the Mask-RCNN (Mask-Resolution Convolutional Neural Networks) model, that is, the region
  • the output of the division layer is the output of the Mask-RCNN model through the mask layer (mask)
  • the output of the region recognition layer is the output of the Mask-RCNN network through the classification layer (class).
  • the motion tracking data of several pedestrian objects contained in the target detection area is determined based on the human body detection images corresponding to the pedestrian images in multiple frames.
  • the motion tracking data includes coordinate information of the pedestrian object in each frame of pedestrian images. Determining the motion tracking data of the human body detection image corresponding to one frame of pedestrian image is described as an example.
  • the above motion tracking data can be the regional center of gravity of the human body area.
  • the terminal device can set the weight of each pixel in the human body area.
  • the pixel coordinates of the regional center of gravity of the human body region can be determined, and the pixel coordinates refer to the coordinates of the center of gravity of the region in the image pixel coordinate system corresponding to the human image; according to the pixel coordinates of the center of gravity of the region and the The depth value corresponding to the barycenter of the region determines the world coordinates of the barycenter of the region, which refers to the coordinates of the barycenter of the region on the world coordinate system corresponding to the target detection region; specifically, according to the pixel coordinates and depth of the barycenter of the region value and the internal parameters of the depth camera, calculate the camera coordinates of the center of gravity of the area, and the camera coordinates refer to the coordinates of the center of gravity of the area in the camera coordinate system corresponding to the depth camera; according to the camera coordinates of the center of gravity of the area and the depth camera
  • the external parameters that is, the transformation matrix from the camera coordinate system to the world coordinate system
  • the above-mentioned determination of the motion tracking data of several pedestrian objects contained in the target detection area based on the human body detection images corresponding to the pedestrian images in multiple consecutive frames further includes: based on the obtained world coordinates, the first line The regional center of gravity of all human body regions in the human image is matched with the regional center of gravity of all human body regions in the second pedestrian image. If the matching is successful, the human body feature information corresponding to the two successfully matched human body regions will be matched.
  • the first pedestrian image and the second pedestrian image are two continuous frames of pedestrian images, that is, the collection of the first pedestrian image and the second pedestrian image
  • the time is the closest, and the human body feature information corresponding to the human body area may specifically be the outline information of the human body area; the world coordinates of the area gravity center of one of the above-mentioned successfully matched human body areas are identified as the above-mentioned associated pedestrian object in the pedestrian image. Coordinate information.
  • the position of the depth camera does not change during the process of acquiring the pedestrian image set, and the internal parameters and external parameters of the depth camera can be obtained by calibrating the depth camera.
  • the pedestrian flow analysis information includes coordinate information of the pedestrian objects in each frame of the pedestrian image, which is used to represent the location of the pedestrian object in each frame of the pedestrian image, and may also include the pedestrian objects in each frame of the pedestrian image.
  • the walking speed in the frame of pedestrian images, the walking speed can be calculated based on the coordinate information of the pedestrian object in each frame of pedestrian images.
  • generating the pedestrian flow analysis information about the target detection area based on the motion tracking data of the several pedestrian objects may specifically include: taking the motion tracking data of a pedestrian object as an example To illustrate, based on the motion tracking data, the walking speed of the pedestrian object in each frame of pedestrian images is calculated. Specifically, an example pedestrian image is used as an example for description, and three frames of adjacent pedestrian images including the example pedestrian image are used.
  • the motion tracking data including the above coordinate information
  • the walking speed of the pedestrian object in the example pedestrian image is zero, it is recognized that the pedestrian object is in a stop state in the example pedestrian image;
  • the maximum collection time interval difference is identified as the stay time period of the pedestrian object in the target detection area. It should be understood that, at this time, the above-mentioned people flow analysis information includes the staying time period of several pedestrian objects in the target detection area.
  • the people flow analysis information may also include the number of pedestrians in each frame of pedestrian images, which is determined according to the number of pedestrian objects in the pedestrian object; the people flow analysis information may also include the number of people flow in the target detection area, the number of people flow Object refers to the total number of distinct pedestrian objects in the above pedestrian image set.
  • a pedestrian image set can be obtained through a depth camera, and multiple frames of pedestrian images in the pedestrian image set can be imported into a human detection network, thereby determining the motion tracking data of pedestrian objects in the target detection area of the depth camera, and further Generate pedestrian flow analysis information, in which the human detection network is trained through deep learning, which can accurately track pedestrians, and the input of the human detection network is the pedestrian image obtained by the depth camera, that is, the depth image, which not only avoids the need for It is easy to infringe the privacy of pedestrians due to the use of color images in the analysis of people flow, and because the depth image can reflect the depth information, it can solve the problem that the people flow analysis through color images in the prior art can only reflect the information in the plane dimension, and cannot The problem of accurately tracking pedestrians.
  • FIG. 2 shows a schematic diagram of an application scenario provided by an embodiment of the present application.
  • the depth camera shoots at the target detection area; the target detection area contains pedestrian objects; the depth camera obtains information about the target The pedestrian image set in the detection area; and then import the multiple frames of pedestrian images in the pedestrian image set into the human detection network, so that the human detection network outputs the corresponding human detection image based on the pedestrian image, and the human detection image is processed.
  • Analysis is performed to obtain the motion tracking data corresponding to the pedestrian object; according to the motion tracking data of several pedestrian objects in the target detection area, the analysis information about the pedestrian flow in the target detection area is generated.
  • FIG. 3 shows a flow chart of the implementation of the method for analyzing the flow of people provided by the second embodiment of the present application.
  • the crowd flow analysis method S102 provided in this embodiment includes S1021 to S1022 , which are described in detail as follows:
  • importing the pedestrian image into a human body detection network, and outputting the human body detection image corresponding to each frame of the pedestrian image includes:
  • the pedestrian image is imported into a human body detection network, so that the human body detection network divides several human body regions in the pedestrian image based on the depth value of each pixel in the pedestrian image.
  • the human body detection network is a neural network model obtained through deep learning training.
  • the human body detection network is a classification model that classifies each area in the pedestrian image to identify the human body whose category is human For details, refer to the relevant description of the above S102, and details are not repeated here.
  • the human detection network includes a region division layer and a region identification layer; the pedestrian image is imported into the human detection network, so that the human detection network is based on the depth of each pixel in the pedestrian image.
  • the pedestrian image is divided into several human body regions.
  • FIG. 4 shows a schematic diagram of a human body detection network provided by an embodiment of the present application. Referring to Figure 4, import the pedestrian image into the region division layer of the human body detection network, and perform edge detection on the pedestrian image based on the depth value of each pixel in the pedestrian image and an edge detection algorithm to obtain several edge contour lines; The edge contour line divides the pedestrian image to obtain several mask areas in the pedestrian image, and each mask area corresponds to an object.
  • the edge detection algorithm can be a depth difference detection algorithm or a gradient difference detection algorithm. ; Based on the parameters of the region division layer in the human body detection network, identify whether the object (category) corresponding to each mask region is a human body, if the object is a pedestrian object, then identify the mask region of the object as a human body region, and Mark all human body regions in this pedestrian image.
  • S1021 based on the positions of the several human body regions in the pedestrian image, configure different region identifiers for each of the human body regions, and mark the human body region and the region marked corresponding to the human body region.
  • the pedestrian image is used as the human detection image.
  • different area identifiers are configured for the marked human body area, so that the same pedestrian object can be identified between multiple frames of pedestrian images subsequently.
  • different region identifiers are configured for each of the human body regions based on the positions of the several human body regions in the pedestrian image, and the human body region and the human body region are marked with different region identifiers.
  • the pedestrian image after the corresponding area identification is used as the human body detection image, which can be specifically: according to the order of the pedestrian object from left to right and from top to bottom, configure a serial number for each human body area, and the serial number is the above-mentioned area. Identification; identify the pedestrian image marked with the human body region and the region identification corresponding to the human body region as the above-mentioned human body detection image.
  • each mask area is divided based on the depth value of each pixel in the pedestrian object, whether each mask area is a human body area is identified, and each human body area is detected number to obtain the above-mentioned human detection image, so that the same pedestrian object can be identified between multiple frames of pedestrian images subsequently.
  • FIG. 5 shows a flowchart of the implementation of the method for analyzing the flow of people provided by the second embodiment of the present application.
  • the crowd flow analysis method S103 provided in this embodiment includes S501 to S505 , which are described in detail as follows:
  • determining the motion tracking data of several pedestrian objects contained in the target detection area based on the human body detection images corresponding to the multiple frames of continuous pedestrian images including:
  • the above-mentioned determination of the position of the center of gravity of the human body corresponding to the area identifier based on the human body region in the human body detection image may specifically be as follows: take the human body detection image corresponding to a frame of pedestrian image as an example for description , determine the position of the center of gravity of each human body region in the human body detection image, specifically, import the human body region into the human body center of gravity recognition network to obtain the pixel coordinates of the human body center of gravity and the depth value of the human body center of gravity; The pixel coordinates and the depth value of the center of gravity of the human body determine the world coordinates of the center of gravity of the human body area, that is, the position of the center of gravity of the human body.
  • the specific method for determining the world coordinates of the center of gravity of the human body can refer to the above-mentioned steps in S103 for determining the world coordinates of the center of gravity of the region , and will not be repeated here.
  • the pixel coordinates of the human body center of gravity and the depth value of the human body center of gravity of the human body region are obtained, and the pixel coordinates of the center of gravity of the human body are obtained; the average value of the depth values of each pixel in the human body area is identified as the depth value of the center of gravity of the human body.
  • the above-mentioned determination of the center of gravity of the human body region based on the contour information of the human body region may specifically be: by analyzing the contour information of the human body region, at least one human body feature point in the human body region is identified, according to the human body feature point. (eg the head) determines the body center of gravity for that body region. The more human feature points are identified in the human body region, the more accurate the determination of the human body's center of gravity.
  • any two adjacent frames of human body detection images are matched based on the position of the center of gravity of the human body and the human body feature information identified by the region.
  • the matching of two adjacent frames of human body detection images refers to the acquisition of two frames of human body detection images with the closest time interval.
  • the above-mentioned matching of any two adjacent frames of human body detection images based on the position of the center of gravity of the human body and the human body feature information identified by the area specifically refers to: matching the above-mentioned one frame of human body detection images The position of the center of gravity of the human body identified by all the regions in the above-mentioned other frame of the human body detection image is compared one-to-one; If the distance between the positions of the center of gravity of the human body identified by the two regions is smaller than the preset distance threshold, the human body feature information of the human body regions corresponding to the two region identifications is compared, and if the human body features of the human body regions corresponding to the two region identifications are compared If the difference value between the information is less than the preset difference threshold, the two region identifiers are identified as two successfully matched region identifiers;
  • the human body area corresponding to the two area identifiers is the human body area of the same pedestrian object, that is, the identification
  • the pedestrian objects associated with the human body regions corresponding to the two area identifiers are the same pedestrian object.
  • the area identifier that matches the area identifier of the area identifier 1 is queried, and the matched area identifier is associated with the pedestrian object No. 1. And so on, until all human regions in all human detection images are associated with corresponding pedestrian objects.
  • the human body region with the region identification 1 is associated with the same pedestrian object as the human body region with the region identification 2 in the second frame of human detection image
  • the pedestrian object associated with the above two human body regions can be identified as the pedestrian object No. 1, that is, the pedestrian object No. 1 is identified by the region between the first frame of human body detection image and the second frame of human body detection image.
  • Human body area move to the human body area with the area ID 2.
  • the motion tracking data of the pedestrian object is determined according to the human body region associated with the human body detection image of the pedestrian object in each frame.
  • the human body area associated with the human body detection image of each pedestrian object in each frame can be obtained.
  • the human body detection of the pedestrian object in each frame is determined. The specific position of the associated human body region in the image is sufficient.
  • the above-mentioned determining the motion tracking data of the pedestrian object according to the human body area associated with the human body detection image of the pedestrian object in each frame may specifically be: extracting the same pedestrian object in The position of the center of gravity of the human body area associated with each frame of human body detection images is converted into the world coordinates of the pedestrian object at the time of collection of each frame of human body detection images, that is, the location of the pedestrian object in the target detection area is obtained.
  • the movement track within the collection period of the pedestrian image set that is, the above-mentioned movement tracking data of the pedestrian object.
  • the human body area corresponding to the same pedestrian object in each frame of human body detection images is used to determine the motion tracking data of the pedestrian object.
  • the people flow analysis method S502 provided in this embodiment also includes S501, and the details are as follows:
  • the method further includes:
  • the depth camera is calibrated to obtain internal parameters of the depth camera.
  • the depth camera is calibrated to obtain the internal parameters of the depth camera, and the internal parameters can be used to convert the position of the point in the human detection image into the relative position of the point relative to the depth camera.
  • the specific performance is: according to the internal parameter, the pixel coordinates of a point on the pixel coordinate system corresponding to the human detection image can be converted into camera coordinates on the camera coordinate system corresponding to the depth camera.
  • pixel coordinates and camera coordinates of multiple feature points need to be obtained to obtain a conversion model for converting pixel coordinates into camera coordinates.
  • the parameters of the conversion model are the internal parameters of the above depth camera. parameter. It should be understood that when the depth camera is calibrated and the human body detection image is subsequently acquired by the depth camera, the depth camera is fixed.
  • the people flow analysis method S502 provided in this embodiment includes S5021 to S5023, and the details are as follows:
  • the determining the position of the center of gravity of the human body corresponding to the region identifier based on the human body region in the human body detection image includes:
  • the pixel coordinates of the center of gravity of the human body are calculated based on the depth value of each pixel in the human body region.
  • the pixel coordinates of the center of gravity of the human body refer to the coordinates of the center of gravity of the human body corresponding to the popular area on the pixel coordinate system corresponding to the human body detection image.
  • the relevant steps of determining the pixel coordinates of the human body's center of gravity in S502 which will not be repeated here.
  • the pixel coordinates are converted into three-dimensional coordinates based on the internal parameters of the depth camera and the depth value of the center of gravity of the human body in the human body detection image.
  • the three-dimensional coordinates may refer to the camera coordinates of the center of gravity of the human body on the camera coordinate system of the depth camera, then the above-mentioned detection image of the human body based on the internal parameters of the depth camera and the center of gravity of the human body Convert the pixel coordinates into three-dimensional coordinates, refer to the following formula for details:
  • K is the internal parameter of the camera
  • Zc is the depth value of the body's center of gravity in the body detection image
  • (u, v) and (Xc, Yc, Zc) are the pixel coordinates of the body's center of gravity and camera coordinates; It is understood that the camera coordinates (Xc, Yc, Zc) of the body's center of gravity, that is, the above-mentioned three-dimensional coordinates, can be determined by solving the above formula.
  • the three-dimensional coordinates of the body's center of gravity are identified as the position of the body's center of gravity corresponding to the area identifier.
  • the three-dimensional coordinates of the center of gravity of the human body obtained in the above S5022 are the camera coordinates of the center of gravity of the human body on the camera coordinate system of the depth camera, and the position of the center of gravity of the human body in this embodiment may refer to the human body
  • the camera coordinates of the center of gravity on the camera coordinate system of the depth camera may also refer to the world coordinates of the center of gravity of the human body on the world coordinate system, which can be specifically set according to user requirements.
  • the camera coordinates (Xc, Yc, Zc) of the body's center of gravity are identified as the position of the body's center of gravity. If the position of the center of gravity of the human body refers to the world coordinates of the center of gravity of the human body in the world coordinate system, the camera coordinates of the center of gravity of the human body should be converted into world coordinates according to the external parameters of the depth camera.
  • the external parameters of the depth camera can be determined by the The relative position between the depth camera and this world coordinate system is determined.
  • the internal parameters of the depth camera are obtained by calibrating the depth camera, and the pixel coordinates of the body's center of gravity are converted into three-dimensional coordinates through the internal parameters, so as to determine the position of the body's center of gravity.
  • FIG. 6 shows a flowchart of the implementation of the method for analyzing the flow of people provided by the second embodiment of the present application.
  • the crowd flow analysis method provided in this embodiment includes S601 to S603 , which are described in detail as follows:
  • preprocessing the pedestrian image including:
  • the above-mentioned contour extraction is performed on the pedestrian image based on the depth value of each pixel in the pedestrian image to obtain the contour line of the pedestrian image, which may specifically be: based on the depth of each pixel in the pedestrian image. value, and perform edge detection on the pedestrian image.
  • the depth value of each pixel of the pedestrian image is processed based on the Prewitt operator, and the edge pixels of the pedestrian image are identified, and determined and extracted according to all the edge pixels.
  • the contour line is a line formed by a plurality of edge pixels.
  • foreground extraction is performed on the pedestrian image based on the background image and the outline to obtain a foreground area of the pedestrian image.
  • the background image is determined after importing the pedestrian image set into a background modeling algorithm, or acquired based on the depth camera; generally, the background image may be the depth camera at the target It is obtained by shooting when there is no pedestrian in the detection area; in particular, the background image can be obtained by importing multiple pedestrian image sets into the background modeling algorithm to identify each pixel coordinate (used to represent the image) in each pedestrian image. The position of the pixel in the pedestrian image) corresponds to the background pixel, and the background image is constructed based on all the identified background pixels.
  • the above-mentioned foreground extraction is performed on the pedestrian image based on the background image and the outline to obtain the foreground area of the pedestrian image.
  • the background image may be compared with the pedestrian image. , if the depth value difference between the pedestrian image and the same pixel coordinate in the background image is greater than the preset comparison threshold, the pixel corresponding to the pixel coordinate is identified as a foreground pixel; all foreground pixels are enclosed by the contour line into at least one The foreground area, that is, the foreground area of the above pedestrian image is obtained.
  • the foreground region is used for the human body detection network to determine the human body region according to the normalized depth value of each pixel in the foreground region.
  • the pedestrian image is a depth image collected by the depth camera.
  • the depth value of each pixel in the depth image has a wide range.
  • the depth value of each pixel in the pedestrian image can be normalized. Normalization processing, and the actual data that needs to be processed in the subsequent human detection network is only the foreground image of the pedestrian image, so the depth value of each pixel in the foreground area is specially normalized.
  • the preprocessed pedestrian image is obtained by the normalized depth value of the pixels.
  • the depth value of the pixel in the non-foreground area in the preprocessed pedestrian image is zero or empty, so that the preprocessed pedestrian image can be subsequently processed.
  • the human detection network may not need to process the pixels in the non-foreground area.
  • the pedestrian image is preprocessed before the pedestrian image is imported into the human body detection network, in order to reduce the data at the input end of the human body detection network to improve efficiency;
  • the contour line of the pedestrian image is extracted to facilitate the determination of the foreground area of the pedestrian image; only the depth values of each pixel in the foreground area are normalized, and the normalized depth of each pixel in the foreground area is processed.
  • the value of the preprocessed pedestrian image is obtained, so that the subsequent human detection network can process the preprocessed pedestrian image, reduce the input amount, improve the efficiency, and also enable the human detection network to converge faster during training and speed up the training process. It should be understood that when the preprocessed pedestrian image is imported into the human body detection network, the human body detection network only needs to identify the human body region in the foreground region.
  • FIG. 7 shows a flowchart of the implementation of the method for analyzing the flow of people provided by the second embodiment of the present application.
  • the crowd flow analysis method S104 provided in this embodiment includes S1041 to S1042, and the details are as follows:
  • generating the people flow analysis information about the target detection area based on the motion tracking data of the several pedestrian objects includes:
  • S1041 based on the motion tracking data of the several pedestrian objects, respectively determine the pedestrian coordinates and the pedestrian speed of the pedestrian object in the target detection area in each frame of the pedestrian image.
  • the pedestrian coordinates may specifically be the world coordinates of the pedestrian object at the time of collection of each frame of pedestrian images; the pedestrian speed specifically refers to the walking speed of the pedestrian object at the time of collection of the pedestrian image;
  • the motion tracking data is the motion data of the pedestrian object in the target detection area, and records the position of the pedestrian object at the collection moment of each frame of pedestrian images and the motion track during the collection period of the pedestrian image set.
  • the pedestrian coordinates and the pedestrian speed of the pedestrian object in each frame of the pedestrian image in the target detection area are respectively determined, which may be specifically: : Take a pedestrian object as an example to illustrate, obtain the world coordinates of the pedestrian object in each frame of pedestrian images from the motion tracking data of the pedestrian object, identify it as the above-mentioned pedestrian coordinates, and the world coordinates of the pedestrian object in each frame of pedestrian images.
  • the coordinates are specifically obtained in S103, and will not be repeated here; the formula for calculating the pedestrian speed of the pedestrian object in each frame of pedestrian images is as follows:
  • v i is the pedestrian speed of the pedestrian object in the ith frame of the pedestrian image
  • s i s i-1 is the position (x i , y i , z i ) of the pedestrian object in the ith frame of the pedestrian image and the The distance between the positions (x i-1 , y i-1 , z i-1 ) of the i-1th frame of pedestrian images
  • dt is the collection interval for the depth camera to collect pedestrian images. It should be understood that when calculating the pedestrian speed of the pedestrian object in the last frame of the pedestrian image, it is generally based on the distance between the position of the pedestrian object in the last frame of the pedestrian image and the position of the pedestrian image in the previous frame of the last frame of the pedestrian image. distance is calculated.
  • the pedestrian coordinates and the pedestrian speed are imported into a preset people flow analysis template to obtain the people flow analysis information.
  • the human flow analysis template is preset and can be adjusted according to requirements.
  • the pedestrian flow analysis information includes the pedestrian coordinates and the pedestrian speed of the pedestrian object at the time when the pedestrian object is collected in each frame. Exemplarily,
  • the pedestrian flow analysis template can be adjusted according to requirements; the pedestrian flow analysis information can also include the total number of pedestrian objects in the target detection area within the collection period of the pedestrian object, which can be obtained by traversing the pedestrian objects in all pedestrian images. Confirm; the pedestrian flow analysis information may also include abnormally stopped pedestrian objects and their corresponding abnormally stopped positions and abnormally stopped durations, and the abnormally stopped pedestrian objects are the pedestrian coordinates and pedestrian speeds of each pedestrian object in each frame of pedestrian images are imported into the pedestrian flow Pedestrian objects output after analyzing the template, specifically, identifying pedestrian objects in each pedestrian object with a pedestrian speed of zero in consecutive multi-frame pedestrian images and the number of frames in the above-mentioned consecutive multi-frame pedestrian images is greater than a preset threshold, identified as For the abnormal pedestrian object, the position of the abnormal pedestrian object in the pedestrian image with the pedestrian speed of zero is identified as the abnormal stay position, and the abnormal stay duration is calculated based on the number of frames of the continuous multiple frames of pedestrian images.
  • FIG. 8 shows a flow chart of the implementation of the method for analyzing the flow of people provided by the second embodiment of the present application.
  • the method for analyzing the flow of people provided by this embodiment includes S801 to S802 , which are described in detail as follows:
  • the method further includes:
  • the first monitoring schematic diagram corresponding to the pedestrian image in each frame is sequentially displayed, and a monitoring image of the target detection area is generated.
  • the first monitoring schematic diagram includes the pedestrian image and the pedestrian flow analysis information corresponding to the pedestrian image.
  • the above-mentioned first monitoring schematic diagram corresponding to each frame of the pedestrian image is displayed in sequence, and the monitoring screen of the target detection area is generated.
  • the first monitoring schematic diagram corresponding to one frame of pedestrian image may be used.
  • the pedestrian flow analysis information obtained in the above S1042 is displayed on the pedestrian image, and packaged into the above-mentioned first monitoring schematic diagram.
  • FIG. 10 shows the monitoring screen provided by the sixth embodiment of the present application. A schematic diagram of the effect.
  • the monitoring screen displays a frame of a first monitoring schematic diagram.
  • each content in the pedestrian flow analysis information is distributed in the corresponding position in the pedestrian image.
  • the pedestrian coordinates and pedestrian speed of the pedestrian object are marked (it should be understood that the pedestrian number of the pedestrian object can also be marked), and the upper left corner of the pedestrian image is marked with all pedestrian objects in the pedestrian image.
  • the total number; the first monitoring schematic diagram corresponding to all the pedestrian images in the above pedestrian image set is formed into the monitoring screen of the target detection area.
  • the monitoring picture is updated in real time based on the second monitoring schematic diagram corresponding to the newly collected pedestrian image.
  • the above-mentioned newly collected pedestrian image refers to a new pedestrian image obtained through the depth camera after acquiring the pedestrian image set through the depth camera in S101, and the newly collected pedestrian image is determined.
  • the corresponding second monitoring schematic diagram reference may be made to the relevant description of the above-mentioned first monitoring schematic diagram, which will not be repeated here.
  • the purpose of updating the monitoring screen in real time based on the second monitoring schematic diagram corresponding to the newly collected pedestrian image is to maintain real-time monitoring of the target detection area, so as to obtain the pedestrian flow analysis information of the target detection area in each time period.
  • FIG. 11 shows a schematic flowchart of a pedestrian flow analysis method provided by an embodiment of the present application.
  • a pedestrian image set in a target detection area is obtained through the depth camera;
  • the pedestrian images in the image set are preprocessed; by importing the preprocessed pedestrian images into the human body detection network, the human body area in the pedestrian image is detected and marked, and the human body detection image is obtained;
  • the pixel coordinates of each pedestrian object are converted into world coordinates, and the motion tracking data of the pedestrian object is obtained; based on the motion tracking data of the pedestrian object, various attribute characteristics of the pedestrian object (including location and speed), generate the people flow analysis information of the target detection area; based on the people flow analysis information of the target detection area, and the pedestrian image set of the target detection area, generate the monitoring screen of the target detection area to visually display the target detection area flow of people.
  • analysis information of the flow of people in the target detection area can be uploaded to the server, so as to facilitate statistics on the flow of people in the target detection area.
  • FIG. 12 shows a schematic structural diagram of a human flow analysis device provided by an embodiment of the present application. For convenience of description, only parts related to the embodiments of the present application are shown.
  • the pedestrian flow analysis device includes: a pedestrian image acquisition module for acquiring a pedestrian image set in the target detection area through a depth camera; the pedestrian image set includes multiple frames of pedestrian images; a human body detection network module for The pedestrian image is imported into the human body detection network, and the human body detection image corresponding to each frame of the pedestrian image is output; the human body detection network is obtained through deep learning training; the motion tracking data determination module is used for the pedestrian image corresponding to the continuous multiple frames.
  • the human body detection image determines the motion tracking data of several pedestrian objects contained in the target detection area; the human flow analysis information generation module is configured to generate information about the target detection based on the motion tracking data of the several pedestrian objects People flow analysis information in the area.
  • FIG. 12 shows a schematic structural diagram of a terminal device provided by an embodiment of the present application.
  • the terminal device 12 in this embodiment includes: at least one processor 120 (only one processor is shown in FIG. 12 ), a memory 121 , and a memory 121 that is stored in the memory 121 and can be processed in the at least one processor
  • a computer program 122 running on the processor 120, and the processor 120 implements the steps in any of the foregoing method embodiments when the computer program 122 is executed.
  • Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the foregoing method embodiments can be implemented.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium.
  • the present application realizes all or part of the processes in the methods of the above embodiments, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When executed by a processor, the steps of each of the above method embodiments can be implemented.
  • the disclosed apparatus/terminal device and method may be implemented in other manners.
  • the apparatus/terminal device embodiments described above are only illustrative.
  • the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units. Or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Multimedia (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Human Computer Interaction (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Image Analysis (AREA)

Abstract

Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur, applicables au domaine de la vision artificielle. Le procédé consiste à : acquérir un ensemble d'images de piétons concernant une région de détection cible au moyen d'un dispositif de prise vues de profondeur (S101), l'ensemble d'images de piétons comprenant de multiples trames d'images de piétons ; importer les images de piétons dans un réseau de détection humaine, et délivrer en sortie une image de détection humaine correspondant à chacune des multiples trames d'images de piéton (S102) ; déterminer, sur la base des images de détection humaine correspondant à de multiples trames consécutives d'images de piéton, des données de suivi de mouvement d'une pluralité d'objets de piéton présents dans la région de détection cible (S103) ; et générer des informations d'analyse de flux de piétons concernant la région de détection cible sur la base des données de suivi de mouvement de la pluralité d'objets de piéton (S104). Le réseau de détection humaine est entraîné au moyen d'un apprentissage profond, un piéton peut être suivi avec précision, et l'image de piéton est une carte de profondeur, de telle sorte que le problème selon lequel une image couleur peut violer la vie privée du piéton est évité et des informations, à l'exception de celles dans une dimension en plan, sont également affichées.
PCT/CN2021/107938 2021-03-19 2021-07-22 Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur WO2022193516A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110297860.6 2021-03-19
CN202110297860.6A CN113034544B (zh) 2021-03-19 2021-03-19 一种基于深度相机的人流分析方法及装置

Publications (1)

Publication Number Publication Date
WO2022193516A1 true WO2022193516A1 (fr) 2022-09-22

Family

ID=76472084

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/107938 WO2022193516A1 (fr) 2021-03-19 2021-07-22 Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur

Country Status (2)

Country Link
CN (1) CN113034544B (fr)
WO (1) WO2022193516A1 (fr)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113034544B (zh) * 2021-03-19 2024-10-15 奥比中光科技集团股份有限公司 一种基于深度相机的人流分析方法及装置
CN115131828A (zh) * 2022-06-29 2022-09-30 珠海视熙科技有限公司 客流人体检测方法、装置、存储介质及客流统计相机

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108986064A (zh) * 2017-05-31 2018-12-11 杭州海康威视数字技术股份有限公司 一种人流量统计方法、设备及系统
US20190378287A1 (en) * 2018-06-12 2019-12-12 Chiral Software, Inc. System and method for object recognition using depth mapping
CN110717408A (zh) * 2019-09-20 2020-01-21 台州智必安科技有限责任公司 一种基于tof相机的人流计数方法
CN111192293A (zh) * 2019-12-27 2020-05-22 深圳市越疆科技有限公司 一种运动目标位姿跟踪方法及装置
CN111709974A (zh) * 2020-06-22 2020-09-25 苏宁云计算有限公司 基于rgb-d图像的人体跟踪方法及装置
CN113034544A (zh) * 2021-03-19 2021-06-25 奥比中光科技集团股份有限公司 一种基于深度相机的人流分析方法及装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101325690A (zh) * 2007-06-12 2008-12-17 上海正电科技发展有限公司 监控视频流中人流分析与人群聚集过程的检测方法及系统
CN104835147A (zh) * 2015-04-15 2015-08-12 中国科学院上海微系统与信息技术研究所 基于三维深度图数据的密集人流量实时检测方法
WO2017000115A1 (fr) * 2015-06-29 2017-01-05 北京旷视科技有限公司 Procédé et dispositif de réidentification de personne
CN107563347B (zh) * 2017-09-20 2019-12-13 南京行者易智能交通科技有限公司 一种基于tof相机的客流计数方法和装置
CN108875588B (zh) * 2018-05-25 2022-04-15 武汉大学 基于深度学习的跨摄像头行人检测跟踪方法
CN109344690B (zh) * 2018-08-09 2022-09-23 上海青识智能科技有限公司 一种基于深度相机的人数统计方法
CN108985265A (zh) * 2018-08-10 2018-12-11 南京华捷艾米软件科技有限公司 客流量监控系统和客流量监控方法
CN109902551A (zh) * 2018-11-09 2019-06-18 阿里巴巴集团控股有限公司 开放场景的实时人流统计方法和装置
CN111753638B (zh) * 2020-05-03 2024-07-16 奥比中光科技集团股份有限公司 一种基于rgbd图像的行人跟踪方法及系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108986064A (zh) * 2017-05-31 2018-12-11 杭州海康威视数字技术股份有限公司 一种人流量统计方法、设备及系统
US20190378287A1 (en) * 2018-06-12 2019-12-12 Chiral Software, Inc. System and method for object recognition using depth mapping
CN110717408A (zh) * 2019-09-20 2020-01-21 台州智必安科技有限责任公司 一种基于tof相机的人流计数方法
CN111192293A (zh) * 2019-12-27 2020-05-22 深圳市越疆科技有限公司 一种运动目标位姿跟踪方法及装置
CN111709974A (zh) * 2020-06-22 2020-09-25 苏宁云计算有限公司 基于rgb-d图像的人体跟踪方法及装置
CN113034544A (zh) * 2021-03-19 2021-06-25 奥比中光科技集团股份有限公司 一种基于深度相机的人流分析方法及装置

Also Published As

Publication number Publication date
CN113034544B (zh) 2024-10-15
CN113034544A (zh) 2021-06-25

Similar Documents

Publication Publication Date Title
WO2021082635A1 (fr) Procédé et appareil de détection de zone d'intérêt, support de stockage lisible et dispositif terminal
US11928800B2 (en) Image coordinate system transformation method and apparatus, device, and storage medium
CN110147717B (zh) 一种人体动作的识别方法及设备
US20220254105A1 (en) Systems and Methods for 3D Facial Modeling
WO2019128507A1 (fr) Procédé et appareil de traitement d'image, support d'informations et dispositif électronique
WO2019128508A1 (fr) Procédé et appareil de traitement d'image, support de mémoire et dispositif électronique
WO2017190646A1 (fr) Procédé et appareil de traitement d'image de visage et support d'informations
WO2022193516A1 (fr) Procédé et appareil d'analyse de flux de piétons basée sur un dispositif de prise vues de profondeur
JP5569990B2 (ja) 属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム
WO2021027537A1 (fr) Procédé et appareil pour prendre une photo d'identification, dispositif et support de stockage
WO2018141252A1 (fr) Procédé de suivi facial, appareil, support d'informations et dispositif électronique
CN109948420B (zh) 人脸比对方法、装置及终端设备
US11850747B2 (en) Action imitation method and robot and computer readable medium using the same
CN105868574A (zh) 一种摄像头跟踪人脸的优化方法及基于视频的智慧健康监视系统
CN110969644A (zh) 人员轨迹追踪方法、装置及系统
CN110472460A (zh) 人脸图像处理方法及装置
CN105022999A (zh) 一种人码伴随实时采集系统
CN104091173B (zh) 一种基于网络摄像机的性别识别方法及装置
TWI731503B (zh) 活體臉部辨識系統與方法
WO2020164266A1 (fr) Procédé et système de détection de corps vivant, et dispositif terminal
WO2023000253A1 (fr) Procédé et appareil d'avertissement précoce de comportement d'escalade, dispositif d'électrode et support de stockage
WO2019223763A1 (fr) Procédé, dispositif et équipement de détection d'image, support, système de commande de formation de motifs et procédé
CN110910449B (zh) 识别物体三维位置的方法和系统
CN110738607A (zh) 基于人工智能拍摄驾驶证的方法、装置、设备及存储介质
CN110321782B (zh) 一种检测人体特征信号的系统

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: 21931094

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: 21931094

Country of ref document: EP

Kind code of ref document: A1