A gait big data retrieval method and its system and terminal device

Abstract

The present invention discloses a gait big data retrieval method and its system and terminal device, which belong to the field of gait recognition technology, and can enable users to use gait retrieval technology in a standardized and correct manner, obtain the best retrieval effect, and thus draw the target person's trajectory information based on the retrieval results in a richer and more accurate manner, solving the problem that the retrieval effect varies from person to person. The method comprises: S1, obtaining a video image containing a target person, extracting a gait sample of the target person from the video image, and constructing a gait library of the target person based on the gait sample; S2, using the gait library to search in the surveillance video data of a local area to obtain a local retrieval result, and using the local retrieval result to update the gait library to obtain an updated gait library; S3, using the updated gait library to search in the surveillance video data of the entire domain to obtain a global retrieval result; S4, generating the trajectory of the target person based on the local retrieval result and the global retrieval result. The present invention is used for gait retrieval.

Description

Technical Field

[0001] The present invention relates to a gait big data retrieval method and a system and terminal equipment thereof, belonging to the technical field of gait recognition. Background Art

[0002] In the AI ​​era, more and more biometric features can be extracted and identified. Currently, the technologies that are more widely used in the market include face recognition, fingerprint recognition, voiceprint recognition, iris recognition, palm vein recognition, and gait recognition.

[0003] Gait recognition refers to the identification of a target person by analyzing his / her body structure and walking posture. Compared with other biometric recognition technologies, gait recognition has the advantages of non-contact, long-distance and not easy to disguise. It is widely used in the fields of safe cities and social governance, such as tracking of missing children, dynamic control of target suspects, and dynamic monitoring of key groups.

[0004] As a new biometric technology in the current security industry, gait recognition has achieved a high recognition accuracy in academia. However, in actual applications, many users have a large number of missing search results or mixed with noise data when searching for data based on gait recognition technology. This problem of inconsistent search results caused by insufficient user experience or improper operation has become increasingly serious as more applications are implemented. Summary of the invention

[0005] The present invention provides a gait big data retrieval method and its system, and terminal device, which can enable users to use gait retrieval technology in a standardized and correct manner to obtain the best retrieval effect, so that target person trajectory information can be drawn more richly and accurately based on the retrieval results, solving the problem that the retrieval effect varies from person to person.

[0006] In one aspect, the present invention provides a gait big data retrieval method, the method comprising:

[0007] S1, obtaining a video image containing a target person, extracting a gait sample of the target person from the video image, and constructing a gait library of the target person according to the gait sample;

[0008] S2, using the gait library to search in the surveillance video data of a local area to obtain a local search result, and using the local search result to update the gait library to obtain an updated gait library;

[0009] S3, using the updated gait library to search in the global surveillance video data to obtain a global search result;

[0010] S4. Generate a target person's action trajectory according to the local search result and the global search result.

[0011] Optionally, the S2 specifically includes:

[0012] S21, obtaining the initial time and place where the target person appears as the center point;

[0013] S22, using the gait library to search the surveillance video data within a preset range around the center point to obtain a local search result;

[0014] S23, updating the gait library using the local search result to obtain an updated gait library;

[0015] S24, taking the time and place where the target person newly appears in the local search result as the updated center point;

[0016] S25, repeatedly executing S22 to S24 until no new track information of the target person is found in the local search results or the target person is found to have left the local search area.

[0017] Optionally, after S3, the method further includes:

[0018] S5, determining whether the newly added trajectory information of the target person is retrieved in the global search results; if so, executing step S6, if not, executing step S4;

[0019] S6. Use the time and place in the newly added trajectory information of the target person in the global search result as the updated center point, and execute step S22.

[0020] Optionally, the S23 specifically includes:

[0021] The gait samples that meet the preset rules in the local search results are added to the gait library as excellent gait samples to obtain an updated gait library.

[0022] Optionally, the surveillance video data within a preset range around the center point are all surveillance video data within a range centered at the center point, radiating 3 kilometers outward, and extending 30 minutes before and after.

[0023] Optionally, the S22 is specifically:

[0024] S221, taking the gait library as a comparison object, performing gait comparison with the surveillance video data within a preset range around the center point, and obtaining a comparison similarity of each surveillance video data;

[0025] S222: Take the gait samples whose comparison similarity is greater than or equal to a preset threshold and their corresponding time and place as local search results.

[0026] Optionally, the S4 is specifically:

[0027] All the time and places where the target person appears in the local search results and the global search results are sorted by time to form a movement track of the target person.

[0028] In another aspect, the present invention provides a gait big data retrieval system, the system comprising:

[0029] A gait library construction unit is used to obtain a video image containing a target person, extract a gait sample of the target person from the video image, and construct a gait library of the target person according to the gait sample;

[0030] A local retrieval unit, used to use the gait library to search in the surveillance video data of a local area to obtain a local retrieval result, and use the local retrieval result to update the gait library to obtain an updated gait library;

[0031] A global search unit, used to use the updated gait library to search in the global surveillance video data to obtain a global search result;

[0032] The trajectory generating unit is used to generate the action trajectory of the target person according to the local search result and the global search result.

[0033] Optionally, the system further includes a data query unit and a data synchronization unit;

[0034] The data query unit is used to query the data in the gait database according to any one or any combination of database type, creation time, gait sample level, and identity information of the collected person;

[0035] The data synchronization unit is used to synchronize the data in the gait library to a superior system or a third-party system according to a preset synchronization period or a preset synchronization strategy.

[0036] In another aspect, the present invention provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor;

[0037] When the processor executes the computer program, the steps of any one of the above methods are implemented.

[0038] The beneficial effects that the present invention can produce include:

[0039] (1) The gait big data retrieval method provided by the present invention can dynamically build a database when the gait samples are insufficient, enrich the number of gait samples of the target person, and improve the quality of the gait samples of the target person. For example, when the pixel of the registered gait sample of the target person is low and the human figure outline is blurred, other excellent gait samples can be quickly found through low-threshold local retrieval.

[0040] (2) The gait big data retrieval method provided by the present invention can quickly obtain more time and place information of the target person's appearance through global retrieval, and can avoid missing the trajectory of the target person in a local range through local retrieval, so as to finely draw the movement trajectory of the target person in a large range. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a gait big data retrieval method provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of the principle of a gait big data retrieval method provided by an embodiment of the present invention;

[0043] Figure 3 A flow chart of a gait big data retrieval method provided by another embodiment of the present invention;

[0044] Figure 4 A schematic diagram of the helper search operation process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with embodiments, but the present invention is not limited to these embodiments.

[0046] The embodiment of the present invention provides a gait big data retrieval method, such as Figure 1 and Figure 2 As shown, the method includes:

[0047] S1. Obtain a video image containing a target person, extract a gait sample of the target person from the video image, and construct a gait library of the target person according to the gait sample.

[0048] Among them, the gait sample can be one or more continuous gait image sequences containing gait features; the gait library can also be called the gait base library; the function of step S1 is to register the base library. Specifically, a video containing the target person's information is uploaded to the system, and the gait sample of the target person is extracted through the gait algorithm to form the gait library of the target person.

[0049] S2. Search the surveillance video data of the local area using the gait library to obtain a local search result, and update the gait library using the local search result to obtain an updated gait library.

[0050] The S2 specifically includes:

[0051] S21, obtaining the initial time and place where the target person appears as the center point. The initial time and place where the target person appears can be the time and place where the target person appears with high probability inferred by the public security personnel based on experience, or can be the time and place inferred based on some relevant evidence materials. The embodiment of the present invention does not limit this.

[0052] S22. Search the surveillance video data within a preset range around the center point using the gait library to obtain a local search result.

[0053] The S22 specifically includes:

[0054] S221. Use the gait library as a comparison object, and perform gait comparison with the surveillance video data within a preset range around the center point to obtain the comparison similarity of each surveillance video data.

[0055] S222: Take the gait samples whose comparison similarity is greater than or equal to a preset threshold and their corresponding time and place as local search results.

[0056] The preset threshold is a value pre-set by those skilled in the art according to actual conditions, and the embodiment of the present invention does not limit this. For example, it can be 80%, 85%, 90%, etc.

[0057] The preset range is a preset time and geographical range, which can be set by a person skilled in the art or a user according to actual conditions, and the embodiments of the present invention do not limit this. For example, the surveillance video data within the preset range around the center point can be all surveillance video data within a range of 3 kilometers radiating outward from the center point and extending 30 minutes before and after the time.

[0058] In practical applications, step S22 can also be called helper retrieval; taking the time and place where the target person has a high probability of appearing as the center point, radiating 3 kilometers outward from the place, and extending 30 minutes before and after the time, select relevant monitoring video data, use the target person's gait library as the comparison object, set the gait comparison threshold to 80%, set the comparison results to be sorted from high to low according to the gait comparison similarity, and perform gait retrieval under the above parameter configuration. The above function can be fixed in the software system to achieve one-click retrieval.

[0059] S23. Update the gait library using the local search results to obtain an updated gait library.

[0060] Specifically: gait samples that meet preset rules in the local retrieval results are added to the gait library as excellent gait samples to obtain an updated gait library.

[0061] The preset rules are pre-set judgment rules, which can be set by those skilled in the art according to actual conditions, and are not limited in the embodiments of the present invention. For example, the preset rules for selecting excellent gait samples can be that the gait samples show the natural walking state of a complete human figure, the gait sample pixels are greater than or equal to 60*100, the gait sample frames are greater than or equal to 15, the gait sample contains more than 2 gait cycles, and the human figure contour boundary in the gait sample is clear.

[0062] S24. Use the time and place where the target person newly appears in the local search results as the updated center point.

[0063] S25, repeatedly executing S22 to S24 until no new track information of the target person is found in the local search results or the target person is found to have left the local search area.

[0064] Taking the time and place where the target person has a high probability of appearing as the center point, select all relevant surveillance video data for helper search, select excellent gait samples from the search results and add them to the gait library of the target person, take the time and place where the target person appears in the search results as the new center point, take the updated target person gait library as the new target comparison object, and select all relevant surveillance video data for helper search again. When there are no new search results in the local area through the helper search (that is, no new trajectory information) or the search results show that the target person has left by means of transportation, the local search is determined to be over.

[0065] S3. Use the updated gait library to search in the global surveillance video data to obtain a global search result; wherein the global surveillance video data is all relevant surveillance video data.

[0066] Furthermore, after S3, the method further includes:

[0067] S5. Determine whether the newly added trajectory information of the target person is retrieved in the global search results; if so, execute step S6; if not, execute step S4.

[0068] S6. Use the time and location in the newly added trajectory information of the target person in the global search results as the updated center point, and execute step S22.

[0069] The updated gait library is used as the target comparison object, and all relevant surveillance video data are selected for gait retrieval. The new time and place where the target person appears in the retrieval results are used as the entry point for further local retrieval.

[0070] S4. Generate the target person's action trajectory based on the local retrieval results and the global retrieval results.

[0071] Specifically: sort all the time and places where the target person appears in the local search results and the global search results by time to form the action trajectory of the target person.

[0072] Another embodiment of the present invention provides a gait big data retrieval method, such as Figure 3 and Figure 4 As shown, the method includes:

[0073] S301, obtaining a video image containing a target person, extracting a gait sample of the target person from the video image, and constructing a gait library of the target person according to the gait sample.

[0074] S302: Obtain the initial time and place where the target person appears as the center point.

[0075] S303: Use the gait library to search in the surveillance video data within a preset range around the center point to obtain a local search result.

[0076] S304: Add the gait samples that meet the preset rules in the local search results as excellent gait samples to the gait library to obtain an updated gait library.

[0077] S305, determine whether the local search results contain the time and place where the target person appears newly; if so, execute step 306; if not, execute step 307.

[0078] S306: Use the time and place where the target person newly appears in the local search results as the updated center point, and execute step 303.

[0079] S307: Use the updated gait library to search in the global surveillance video data to obtain a global search result.

[0080] S308, determining whether the global search results include the time and place where the target person newly appears; if so, executing step 309; if not, executing step 310.

[0081] S309: Use the time and place where the target person newly appears in the global search results as the updated center point, and execute step 303.

[0082] S310, sorting all the time and places where the target person appears in the local search results and the global search results by time to form a movement trajectory of the target person.

[0083] The present invention can dynamically build a database when gait samples are insufficient, enrich the number of gait samples of the target person, and improve the quality of the gait samples of the target person. For example, when the registered gait sample of the target person has low pixels and the outline of the human figure is blurred, other excellent gait samples can be quickly found through low-threshold local search. At the same time, more time and place information of the target person's appearance can be quickly obtained through global search, and the trajectory of the target person in the local range can be avoided from being missed through local search, so that the action trajectory of the target person in a large range can be finely drawn.

[0084] Another embodiment of the present invention provides a gait big data retrieval system, the system comprising:

[0085] The gait library construction unit is used to obtain a video image containing a target person, extract a gait sample of the target person from the video image, and construct a gait library of the target person based on the gait sample.

[0086] The local retrieval unit is used to use the gait library to search in the monitoring video data of the local area to obtain the local retrieval result, and use the local retrieval result to update the gait library to obtain the updated gait library.

[0087] The global retrieval unit is used to use the updated gait library to search in the global surveillance video data to obtain the global retrieval result.

[0088] The trajectory generation unit is used to generate the action trajectory of the target person according to the local retrieval results and the global retrieval results.

[0089] Furthermore, the system also includes a data query unit and a data synchronization unit;

[0090] The data query unit is used to query the data in the gait database according to any one or any combination of database type, creation time, gait sample level and identity information of the collected person, so as to browse the information such as the identity of the person in the base database and gait samples.

[0091] The data synchronization unit is used to synchronize the data in the gait library to the superior system or the third-party system according to a preset synchronization cycle or a preset synchronization strategy.

[0092] The retrieval system provided in the embodiment of the present invention has the functions of assistant retrieval, gait retrieval, gait comparison, retrieval result summary, trajectory drawing, gait database quality assessment, etc.

[0093] The retrieval system provided by the embodiment of the present invention has user management, role management, and organization management functions; it has different usage authority management functions for ordinary users and administrators, and has relevant authorization mechanisms in operations such as gait registration, deletion, and query. At the same time, the retrieval system of the present invention also has a log management function. For each event, the log record should include the time of event occurrence, event type, user, event execution result or failure reason, log validity period, etc.

[0094] Yet another embodiment of the present invention provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the computer program.

[0095] The above are only a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application is disclosed as above with preferred embodiments, it is not intended to limit the present application. Any technician familiar with the profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A gait big data retrieval method, It is characterized in that The method comprises: S1, obtaining a video image containing a target person, extracting a gait sample of the target person from the video image, and constructing a gait library of the target person according to the gait sample; S2, using the gait library to search in the surveillance video data of a local area to obtain a local search result, and using the local search result to update the gait library to obtain an updated gait library; S3, using the updated gait library to search in the global surveillance video data to obtain a global search result; S4. Generate a target person's action trajectory according to the local search result and the global search result.

2. The method according to claim 1, It is characterized in that The S2 specifically includes: S21, obtaining the initial time and place where the target person appears as the center point; S22, using the gait library to search the surveillance video data within a preset range around the center point to obtain a local search result; S23, updating the gait library using the local search result to obtain an updated gait library; S24, taking the time and place where the target person newly appears in the local search result as the updated center point; S25, repeatedly executing S22 to S24 until no new track information of the target person is found in the local search results or the target person is found to have left the local search area.

3. The method according to claim 2, It is characterized in that After S3, the method further includes: S5, determining whether the newly added trajectory information of the target person is retrieved in the global search results; if so, executing step S6, if not, executing step S4; S6. Use the time and place in the newly added trajectory information of the target person in the global search result as the updated center point, and execute step S22.

4. The method according to claim 2, It is characterized in that The S23 specifically includes: The gait samples that meet the preset rules in the local search results are added to the gait library as excellent gait samples to obtain an updated gait library.

5. The method according to claim 2, It is characterized in that The surveillance video data within a preset range around the center point refers to all surveillance video data within a range centered at the center point, radiating 3 kilometers outward, and extending 30 minutes before and after.

6. The method according to claim 2, It is characterized in that The S22 is specifically: S221, taking the gait library as a comparison object, performing gait comparison with the surveillance video data within a preset range around the center point, and obtaining a comparison similarity of each surveillance video data; S222: Take the gait samples whose comparison similarity is greater than or equal to a preset threshold and their corresponding time and place as local search results.

7. The method according to claim 1, It is characterized in that The S4 is specifically: All the time and places where the target person appears in the local search results and the global search results are sorted by time to form a movement track of the target person.

8. A gait big data retrieval system, It is characterized in that The system comprises: A gait library construction unit is used to obtain a video image containing a target person, extract a gait sample of the target person from the video image, and construct a gait library of the target person according to the gait sample; A local retrieval unit, used to use the gait library to search in the surveillance video data of a local area to obtain a local retrieval result, and use the local retrieval result to update the gait library to obtain an updated gait library; A global search unit, used to use the updated gait library to search in the global surveillance video data to obtain a global search result; The trajectory generating unit is used to generate the action trajectory of the target person according to the local search result and the global search result.

9. The system according to claim 8, It is characterized in that The system also includes a data query unit and a data synchronization unit; The data query unit is used to query the data in the gait database according to any one or any combination of database type, creation time, gait sample level, and identity information of the collected person; The data synchronization unit is used to synchronize the data in the gait library to a superior system or a third-party system according to a preset synchronization period or a preset synchronization strategy.

10. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images