Face and body association method, system and computer-readable storage medium

By clustering and merging face and body images in video streams, the problem of errors in human face and body association in monitoring scenes is solved, and the accuracy and efficiency of the monitoring system are improved.

CN113657434BActive Publication Date: 2024-12-06ZHEJIANG DAHUA TECH CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202110751350.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-07-02
Publication Date
2024-12-06
Estimated Expiration
2041-07-02

AI Technical Summary

Technical Problem

In surveillance scenarios with dense crowds, existing human-body association methods are prone to errors in complex situations, especially when the target is temporarily blocked, resulting in multiple groups of association relationships for the same target, which has low accuracy and efficiency.

Method used

By obtaining the initial image collection from the video stream, clustering face images, combining face and body images with the same tracking logo, and filtering out the face and body images with the highest scores, establishing associations, and avoiding temporary occlusion to cause multiple groups of associations.

Benefits of technology

It improves the accuracy and efficiency of human-body associations, ensures accurate identification of targets, and avoids multiple groups of associations caused by occlusion and other reasons.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113657434B_ABST
    Figure CN113657434B_ABST
Patent Text Reader

Abstract

The present application discloses a face-body association method, system and computer-readable storage medium, the method comprising: obtaining an initial image collection from a captured video stream, the initial image collection comprising a plurality of face images and a plurality of body images, each face image and body image having a corresponding tracking identifier; clustering the plurality of face images to obtain at least one first face image collection; wherein the face images in the same first face image collection have the same or different tracking identifiers; merging the face images and body images with the same tracking identifier in the initial image collection according to the first face image collection, and screening out the face image with the highest score and the body image with the highest score; establishing an association relationship between the face image with the highest score and the body image with the highest score. Avoiding the situation where the same target generates multiple groups of association relationships due to the temporary loss of the target, thereby improving the accuracy and efficiency of face-body association.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a face and body association method, system, and computer-readable storage medium. Background Art

[0002] With the development of artificial intelligence technology, image acquisition is used for monitoring scenarios, including capturing targets at entry and exit ports, railway stations, airport halls, etc. In these crowded scenes, most of the images contain multiple faces and bodies. How to associate faces with corresponding bodies to accurately identify targets is a technical problem that needs to be solved.

[0003] The current method mainly associates a face with the corresponding human body through the relative relationship between the face and the human body. However, during the long-term research and development process, the inventor found that this method has at least the following problems: when the monitored scene is complex, such as overlapping human bodies, the association between the face and the human body is prone to errors; when the target is temporarily blocked, multiple snapshots will appear, and the same target will have multiple groups of associations, and the association between the face and the human body is prone to errors. Therefore, the accuracy and efficiency of the existing face and human body association method are low. Summary of the invention

[0004] The main technical problem solved by the present application is to provide a face and body association method, system and computer-readable storage medium, which can avoid the situation where multiple groups of association relationships are generated for the same target due to the temporary loss of the target.

[0005] To solve the above technical problems, a technical solution adopted in the present application is: to provide a face and body association method, comprising: obtaining an initial image collection from a captured video stream, wherein the initial image collection includes multiple face images and multiple body images, and each of the face image and the body image has a corresponding tracking identifier; clustering the multiple face images to obtain at least one first face image collection; wherein the face images in the same first face image collection have the same or different tracking identifiers; according to the first face image collection, merging the face images and body images with the same tracking identifier in the initial image collection, and screening out the face image with the highest score and the body image with the highest score; establishing an association relationship between the face image with the highest score and the body image with the highest score.

[0006] Among them, the step of clustering the multiple facial images to obtain at least one first facial image collection includes: comparing the corresponding facial features extracted from each two of the facial images to obtain the similarity between each two of the facial images; clustering the facial images whose similarity is greater than a first preset threshold to obtain at least one first facial image collection.

[0007] Among them, the step of comparing the corresponding facial features extracted from each two of the facial images to obtain the similarity between each two of the facial images includes: screening out all facial images whose face width is greater than a second preset threshold and whose face quality score is greater than a third preset threshold from the multiple facial images; and comparing the corresponding facial features extracted from each two of the facial images from the screened facial images to obtain the similarity between each two of the facial images.

[0008] Among them, the step of merging the facial images and human body images with the same tracking identifier in the initial image collection according to the first facial image collection includes: obtaining the tracking identifier of all facial images in the first facial image collection; screening out human body images with the same tracking identifier as the facial images from all the human body images; and merging the screened human body images with their corresponding facial images.

[0009] Among them, the step of obtaining an initial image collection from the acquired video stream, wherein the initial image collection includes face images and body images, and each of the face images and the body images has a corresponding tracking identifier, comprises: in response to detecting the presence of a target in the video stream, tracking the target and determining whether the target disappears; in response to detecting the disappearance of the target, obtaining from the video stream the best body image and its corresponding face image, the best face image and its corresponding body image, and the face image and body image with the highest comprehensive score of face and body within the tracking period, so as to generate the initial image collection.

[0010] Among them, in response to detecting that the target disappears, the step of obtaining the optimal human body image and its corresponding facial image, the optimal facial image and its corresponding human body image, and the facial image and human body image with the highest comprehensive score of face and body in the tracking period in the video stream to generate the initial image collection includes: extracting facial features of all the facial images in the initial image collection; caching the facial features, the facial images and the human body images in the initial image collection, and the corresponding tracking identifiers.

[0011] Among them, after the step of caching the facial features, the facial images and the body images in the initial image collection and the corresponding tracking identifiers, it includes: determining whether the caching time is greater than a fourth preset threshold; if so, entering the step of clustering the facial images to obtain at least one first facial image collection; otherwise, entering the step of tracking the target and determining whether the target disappears in response to detecting the presence of the target in the video stream.

[0012] Among them, the step of establishing an association relationship between the facial image with the highest score and the human body image with the highest score includes: configuring the same label for the facial image with the highest score and the human body image with the highest score; and establishing an association relationship between the facial image with the highest score and the human body image with the highest score based on the label.

[0013] To solve the above technical problems, another technical solution adopted in the present application is: to provide a face-body association system, comprising a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the face-body association method mentioned in any of the above embodiments.

[0014] In order to solve the above technical problems, another technical solution adopted in the present application is: providing a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is used to implement the face-body association method mentioned in any of the above embodiments.

[0015] Different from the prior art, the beneficial effects of the present application are as follows: in the present application, an initial image collection is obtained from the acquired video stream, and multiple face images are clustered to obtain at least one first face image collection, wherein the face images in the same first face image collection have the same or different tracking identifiers, and the face images and body images with the same tracking identifier in the initial image collection are merged according to the first face image collection, and the face images with the highest scores and the body images with the highest scores are screened out, and an association relationship is established between the face images with the highest scores and the body images with the highest scores. Through this design method, face images and body images with the same tracking identifier can be merged according to the clustering results, and the face images and body images with the best quality can be associated according to the merged results, which can avoid the situation where multiple groups of association relationships are generated for the same target due to short-term tracking failures due to occlusion and other reasons, thereby improving the accuracy and efficiency of face-body association. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0017] Figure 1 It is a flowchart of an implementation method of the face and human body association method of the present application;

[0018] Figure 2 yes Figure 1A schematic flow chart of an implementation method of step S1 in FIG.

[0019] Figure 3 yes Figure 2 A schematic diagram of a flow chart of an implementation method after step S13;

[0020] Figure 4 yes Figure 3 A schematic diagram of a flow chart of an implementation method after step S131;

[0021] Figure 5 yes Figure 1 A schematic flow chart of an implementation method of step S2 in FIG.

[0022] Figure 6 yes Figure 5 A schematic diagram of a flow chart of an implementation method of step S20;

[0023] Figure 7 yes Figure 1 A schematic flow chart of an implementation method corresponding to step S3;

[0024] Figure 8 yes Figure 1 A schematic flow chart of an implementation method of step S4 in FIG.

[0025] Fig. 9 It is a schematic diagram of the framework of an implementation method of the face and human body association system of the present application;

[0026] Fig.10 This is a schematic diagram of the structure of an implementation method of the face and human body association system of the present application;

[0027] Fig.11 It is a schematic diagram of a framework of an implementation scheme of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] See also Figure 1 , Figure 1 This is a flowchart of an embodiment of the face-body association method of the present application. Specifically, the method includes:

[0030] S1: Obtain an initial image collection from the acquired video stream.

[0031] Specifically, the above-mentioned initial image collection includes multiple face images and multiple body images, each face image and body image has a corresponding tracking identifier, and the face images or body images with the same tracking identifier belong to the same person.

[0032] Specifically, in this embodiment, please refer to Figure 2 , Figure 2 yes Figure 1 Schematic diagram of a flow chart of an implementation method of step S1 in FIG. Specifically, step S1 includes:

[0033] S10: Detect objects in the video stream.

[0034] Specifically, a front-end device such as a camera may be used to collect the video stream, which is not limited in the present application. The collected video stream is analyzed and the target in the video stream is detected.

[0035] S11: Determine whether there is a target in the video stream.

[0036] S12: If yes, track the target and determine whether the target disappears.

[0037] Specifically, if there is a target in the video stream, the target is tracked and it is determined whether the target disappears.

[0038] S13: Otherwise, return to step S10.

[0039] Specifically, if there is no target in the video stream, the process returns to the step of detecting the target in the video stream.

[0040] S14: If yes, obtain from the video stream the best human image and its corresponding face image within the tracking period, the best face image and its corresponding body image, and the face image and body image with the highest comprehensive score of face and body, to generate an initial image collection.

[0041] Specifically, if the target disappears, the optimal human image and its corresponding face image, the optimal face image and its corresponding body image, and the face image and body image with the highest comprehensive score of face and body are obtained from the video stream during the tracking period to generate an initial image collection. This ensures that the quality of the associated face image and body image is optimal, thereby improving the efficiency and quality of face-body association. In addition, in this embodiment, the tracking identifiers of the acquired face image and body image are saved together, which is convenient for later learning whether the acquired face image and body image belong to the same person and for later establishing an association relationship between the face image and the body image.

[0042] S15: Otherwise, return to the step of tracking the target and determining whether the target disappears.

[0043] Specifically, if the target has not disappeared, the process returns to the step of tracking the target and determining whether the target has disappeared, and continues to track the target until the target disappears.

[0044] Specifically, in this embodiment, please refer to Figure 3 , Figure 3 yes Figure 2 Schematic diagram of a flow chart of an implementation method after step S14. Specifically, after step S14, it includes:

[0045] S140: Extracting facial features of all facial images in the initial image collection.

[0046] Specifically, facial features of all facial images are extracted from the initial image collection obtained in step S14 to cluster all facial images.

[0047] S141: Cache facial features, facial images and body images in the initial image collection, and corresponding tracking identifiers.

[0048] The facial features obtained above, the facial images and body images in the initial image collection, and the corresponding tracking identifiers are saved in the cache queue to facilitate the retrieval of relevant data at any time in the subsequent steps.

[0049] Specifically, in this embodiment, please refer to Figure 4 , Figure 4 yes Figure 3 Schematic diagram of a flow chart of an implementation method after step S141. Specifically, after step S141, it includes:

[0050] S1410: Determine whether the cache time is greater than a fourth preset threshold.

[0051] Specifically, the fourth preset threshold is determined by the user. Optionally, the fourth preset threshold is set to 60 seconds, that is, it is determined whether the cache time is greater than 60 seconds. Of course, in other embodiments, the cache time can also be set to other values, which is not limited in this application.

[0052] S1411: If yes, then enter the step of clustering the facial images to obtain at least one first facial image collection.

[0053] Specifically, if the cache time is greater than the fourth preset threshold, the process proceeds to step S2 to obtain a first facial image collection.

[0054] S1412: Otherwise, return to the step of tracking the target and determining whether the target disappears.

[0055] Specifically, if the cache time is less than or equal to the fourth preset threshold, it means that the cache queue can continue to cache related data, and then the process returns to the step of tracking the target and determining whether the target disappears.

[0056] S2: Clustering multiple face images to obtain at least one first face image collection.

[0057] Specifically, the facial images in the same first facial image collection have the same or different tracking identifiers. The facial images in the first facial image collection obtained after clustering belong to the same category. In this embodiment, the facial images in the first facial image collection obtained after clustering belong to the same target. However, due to the temporary loss of the target, the facial images belonging to the same target may generate two or more different tracking identifiers. The clustering method can be used to merge different files belonging to the same target, thereby improving the accuracy of association.

[0058] Specifically, in this embodiment, please refer to Figure 5 , Figure 5 yes Figure 1 Schematic diagram of a flow chart of an implementation method of step S2 in FIG. Specifically, step S2 includes:

[0059] S20: Compare the corresponding facial features extracted from every two facial images to obtain the similarity between every two facial images.

[0060] Specifically, in this embodiment, please refer to Figure 6 , Figure 6 yes Figure 5 Schematic diagram of a flow chart of an implementation method of step S20 in FIG. Specifically, step S20 includes:

[0061] S200: Filtering out, from a plurality of face images, all face images whose face width is greater than a second preset threshold and whose face quality scores are greater than a third preset threshold.

[0062] Specifically, all face images whose face width is greater than the second preset threshold value X and whose face quality score is greater than the third preset threshold value Y are screened out from multiple face images to ensure that the screened face images are all high-quality images. On the contrary, if the face width of a face image is less than or equal to the second preset threshold value X, the face image is likely to be a case of overlapping human bodies, and association errors are prone to occur. Therefore, if the face width of a face image is less than or equal to the second preset threshold value X, the face image is not clustered; if the face quality score of a face image is less than or equal to the third preset threshold value Y, the face image is not clustered either. This ensures that the quality of the face images in the final clustering results is optimal, thereby improving the accuracy and efficiency of association. In addition, the second preset threshold value X and the third preset threshold value Y can be set according to actual conditions, and this application does not limit them here.

[0063] S201: comparing corresponding facial features extracted from every two facial images among the screened facial images to obtain the similarity between every two facial images.

[0064] Specifically, among the face images screened in the above step S200, the corresponding face features extracted from every two face images are compared to obtain the similarity between every two face images.

[0065] S21: Clustering facial images whose similarity is greater than a first preset threshold to obtain at least one first facial image collection.

[0066] Specifically, face images with a similarity greater than a first preset threshold Z are clustered. Face images in the same first face image collection obtained by clustering are similar and can be considered to belong to the same target. However, due to the temporary loss of the target, the face images belonging to the same target may generate two or more different tracking identifiers. Through clustering, face images belonging to the same target with the same or different tracking identifiers can be merged, so as to obtain the tracking identifiers belonging to the same target, thereby improving the accuracy of association.

[0067] S3: merging the face images and body images with the same tracking identifier in the initial image collection according to the first face image collection, and screening out the face images and body images with the highest scores.

[0068] Specifically, according to the tracking identifiers belonging to the same target in the first face image collection, the face images and body images with the same tracking identifier in the initial image collection are merged, and the face images and body images with the same tracking identifier in the merged result belong to the same target. In this way, the face images and body images belonging to the same target can be merged according to the clustering result, which can avoid the situation where multiple groups of association relationships are generated for the same target due to temporary tracking failures due to occlusion and other reasons, thereby improving the accuracy and efficiency of face-body association. In addition, in this embodiment, the face images and body images with the highest scores belonging to the same target are screened out from the merged results. In this way, the final association result can be the best.

[0069] Specifically, in this embodiment, please refer to Figure 7 , Figure 7 yes Figure 1 Specifically, the step of merging the face images and body images with the same tracking identifier in the initial image collection according to the first face image collection in step S3 includes:

[0070] S30: Obtain tracking identifiers of all face images in the first face image collection.

[0071] Specifically, the tracking identifiers of the facial images in a first facial image collection obtained in the above step S2 may be the same or different, but the facial images in a first facial image collection all belong to the same target.

[0072] S31: Filter out human body images having the same tracking identifier as the face image from all human body images.

[0073] Specifically, human images with the same tracking identifier as the face image are selected from all human images in the initial image collection. The face image and human image with the same identifier obtained in this way must belong to the same target.

[0074] S32: Merging the screened human body image with its corresponding face image.

[0075] Specifically, the merging process is to merge the face images and body images belonging to the same target into one category. In the merged results, the face images and body images with the same tracking identifiers belong to the same target, which can improve the accuracy and efficiency of face-body association.

[0076] S4: Establish an association relationship between the face image with the highest score and the body image with the highest score.

[0077] Specifically, the result obtained in step S3 is a face image and a body image belonging to the same target. Based on the result, the face image with the highest score and the body image with the highest score are associated. The associated faces and bodies are of the best quality and have greater practical value.

[0078] Specifically, in this embodiment, please refer to Figure 8 , Figure 8 yes Figure 1 Schematic diagram of a flow chart of an implementation method of step S4 in FIG. Specifically, step S4 includes:

[0079] S40: Assign the same label to the face image with the highest score and the human image with the highest score.

[0080] Specifically, the face image with the highest score and the body image with the highest score are assigned the same label, so that the face image and the body image can be matched one by one when associating the body and the face in the later stage.

[0081] S41: Establish an association relationship between the face image with the highest score and the body image with the highest score according to the tags.

[0082] Specifically, an association is established between the highest-scoring face image and the highest-scoring body image with the same label.

[0083] Through this design, face images and body images with the same tracking identifier can be merged according to the clustering results, which is convenient for screening out face images and body images belonging to the same target. The face images and body images with the best quality can be associated according to the merged results. This can avoid the situation where multiple groups of associations are generated for the same target due to temporary tracking failures due to occlusion, etc., thereby improving the accuracy and efficiency of face-body association.

[0084] See also Fig. 9 , Fig. 9 This is a schematic diagram of the framework of an embodiment of the face-body association system of the present application. The face-body association system specifically includes:

[0085] The acquisition module 10 is used to acquire an initial image collection from the acquired video stream, wherein the initial image collection includes a plurality of face images and a plurality of body images, and each face image and body image has a corresponding tracking identifier.

[0086] The clustering module 12 is coupled to the acquisition module 10 and is used to cluster the multiple face images to obtain at least one first face image collection; wherein the face images in the same first face image collection have the same or different tracking identifiers.

[0087] The processing module 14 is coupled to the clustering module 12 and is used to merge the face images and body images with the same tracking identifier in the initial image collection according to the first face image collection, and screen out the face images and body images with the highest scores.

[0088] The association module 16 is coupled to the processing module 14 and is used to establish an association relationship between the face image with the highest score and the body image with the highest score.

[0089] See also Fig.10 , Fig.10 1 is a schematic diagram of the structure of an embodiment of the face-body association system of the present application. The face-body association system includes a memory 20 and a processor 22 coupled to each other. Specifically, in this embodiment, the memory 20 stores program instructions, and the processor 22 is used to execute the program instructions to implement the face-body association method mentioned in any of the above embodiments.

[0090] Specifically, the processor 22 may also be referred to as a CPU (Central Processing Unit). The processor 22 may be an integrated circuit chip having the ability to process signals. The processor 22 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. In addition, the processor 22 may be implemented by a plurality of integrated circuit chips.

[0091] See also Fig.11 , Fig.11It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application. The computer-readable storage medium 30 stores a computer program 300, which can be read by a computer, and the computer program 300 can be executed by a processor to implement the face-body association method mentioned in any of the above embodiments. Among them, the computer program 300 can be stored in the above-mentioned computer-readable storage medium 30 in the form of a software product, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The computer-readable storage medium 30 with a storage function can be a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, RandomAccess Memory), a disk or an optical disk, etc., which can store program codes, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0092] In summary, different from the prior art, the present application obtains an initial image collection from the acquired video stream, clusters multiple face images, and obtains at least one first face image collection, wherein the face images in the same first face image collection have the same or different tracking identifiers, and the face images and body images with the same tracking identifier in the initial image collection are merged according to the first face image collection, and the face images with the highest scores and the body images with the highest scores are screened out, and an association relationship is established between the face images with the highest scores and the body images with the highest scores. Through this design method, face images and body images with the same tracking identifier can be merged according to the clustering results, so that face images and body images belonging to the same target can be easily screened out, and face images and body images with the best quality can be associated with each other according to the merged results, which can avoid the situation where multiple groups of association relationships are generated for the same target due to short-term tracking failures due to occlusion and other reasons, thereby improving the accuracy and efficiency of face-body association.

[0093] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A face and body association method, characterized in that: include: In response to detecting the presence of an object in the video stream, tracking the object and determining whether the object disappears; In response to detecting that the target disappears, obtaining from the video stream an optimal human image and its corresponding face image, an optimal face image and its corresponding body image, and a face image and body image with the highest comprehensive face and body scores within a tracking period, so as to generate an initial image collection; wherein the initial image collection includes a plurality of face images and a plurality of body images, each of which has a corresponding tracking identifier; Clustering the plurality of face images to obtain at least one first face image collection; wherein the face images in the same first face image collection have the same or different tracking identifiers; According to the first face image collection, the face images and body images with the same tracking identifier in the initial image collection are merged, and the face images and body images with the highest scores are screened out; An association relationship is established between the face image with the highest score and the body image with the highest score.

2. The face-body association method according to claim 1, characterized in that: The step of clustering the plurality of face images to obtain at least one first face image collection comprises: Comparing the corresponding facial features extracted from each two of the facial images to obtain the similarity between each two of the facial images; The facial images whose similarity is greater than a first preset threshold are clustered to obtain at least one first facial image collection.

3. The face-body association method according to claim 2, characterized in that: The step of comparing the corresponding facial features extracted from each two facial images to obtain the similarity between each two facial images comprises: Filter out all face images whose face width is greater than a second preset threshold and whose face quality score is greater than a third preset threshold from the multiple face images; From the screened face images, the corresponding face features extracted from every two face images are compared to obtain the similarity between every two face images.

4. The face-body association method according to claim 1, characterized in that: The step of merging the face images and body images having the same tracking identifier in the initial image collection according to the first face image collection comprises: Obtaining tracking identifiers of all face images in the first face image collection; Filtering out a human body image having the same tracking identifier as the face image from all the human body images; Merge the filtered human body images with their corresponding face images.

5. The face-body association method according to claim 1, characterized in that: After the step of obtaining, in response to detecting that the target disappears, the best human image and its corresponding face image, the best face image and its corresponding body image, and the face image and body image with the highest comprehensive score of face and body in the tracking period in the video stream to generate the initial image collection, the method further comprises: Extracting facial features of all the facial images in the initial image collection; The facial features, the facial images and the human body images in the initial image collection, and the corresponding tracking identifiers are cached.

6. The face-body association method according to claim 5, characterized in that: After the step of caching the facial features, the facial images and the human body images in the initial image collection, and the corresponding tracking identifiers, the method further includes: Determine whether the cache time is greater than a fourth preset threshold; If yes, then entering the step of clustering the facial images according to the facial features extracted from the facial images to obtain at least one first facial image collection; Otherwise, enter the step of, in response to detecting the existence of a target in the video stream, tracking the target and determining whether the target disappears.

7. The face-body association method according to claim 1, characterized in that: The step of establishing an association relationship between the face image with the highest score and the body image with the highest score comprises: Assigning the same label to the face image with the highest score and the body image with the highest score; An association relationship is established between the face image with the highest score and the body image with the highest score according to the label.

8. A face and body association system, characterized in that: It comprises a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the face-body association method described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is used to implement the face-body association method described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Face image clustering method, device, server and storage medium

    CN109117803A

  • Pedestrian recognition method and device

    CN109753920A

  • Target tracking method, target tracking device and computer readable storage medium

    CN112149762A

  • Full-target association index automatic test method and system, medium and terminal

    CN112733948A

  • Portrait clustering method and device and storage medium

    CN112818867A