Target personnel data processing method and device based on collected data and medium

Abstract

The invention provides a target person data processing method and device based on collected data and a medium, and relates to the technical field of data processing, and the method comprises the steps: collecting the sign data of a target person, so as to obtain a sign feature vector HT; obtaining a target keyword list according to the HT; and obtaining a plurality of key videos according to the target keyword list and the standard behavior description statement corresponding to the target person. Compared with the non-targeted health prompt, the reminding based on accurate data and analysis can better arouse the attention of the personnel and promote the personnel to change bad behaviors.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, device and medium for processing target personnel data based on collected data. Background Technology

[0002] In the field of intelligent security and health monitoring, end-to-end monitoring systems have been widely used in densely populated places such as kindergartens and nursing homes that require close monitoring. However, when monitored individuals exhibit specific physical signs such as myopia or scoliosis, it is urgent to review historical monitoring videos to analyze potential unhealthy behaviors and habits, providing a basis for health intervention and safety management.

[0003] Existing technical solutions suffer from two major bottlenecks: First, manually screening historical surveillance videos is extremely time-consuming and labor-intensive, making it difficult to meet real-time and accuracy requirements. Second, directly inputting complete time-span surveillance video data into a large model for analysis leads to excessive consumption of computing resources due to the large volume and redundant information in video data, resulting in low model training and inference efficiency and failing to meet the rapid response needs of real-world applications. Therefore, how to accurately analyze the causes of standard vital signs and identify undesirable habits while reducing data processing volume has become an urgent technical problem to be solved. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a method, device, and medium for processing target personnel data based on collected data, which at least partially solves the problems existing in the prior art.

[0005] In a first aspect of this application, a method for processing target personnel data based on collected data is provided, the method comprising:

[0006] Collect vital sign data of the target personnel to obtain the vital sign feature vector HT;

[0007] Based on HT, obtain the target keyword list; where the target keyword list is the keyword list corresponding to the target standard vital sign feature vector MT determined by HT from several standard vital sign feature vectors; the matching degree P between MT and HT. MH The matching degree between HT and any standard trait vector other than MT is greater than that between HT and P. MH The match degree is greater than the preset matching threshold; the keywords in the keyword list are used to describe the causes of the standard physical characteristics corresponding to the standard physical characteristic feature vector; the keywords are any standard statements in the standard statement library;

[0008] According to the target keyword list and the standard behavior description statement corresponding to the target person, a plurality of key videos are obtained; the behavior description statement is obtained by standardizing output of key images collected by an image acquisition device in a target area where the target person is located through a large language model; the image acquisition device collects continuous videos at the same angle while collecting images; each key image has a corresponding collection time; the key video is a video corresponding to a preset time length before and after the collection time of the key image by the same image acquisition device; the key image is obtained by classifying the scene type of all images collected by the image acquisition device in a target time window according to the large language model; the key video is used for analyzing and reminding the behavior of the target person; and a time period corresponding to each key video is within the target time window.

[0009] In a second aspect of the present application, a non-transitory computer readable storage medium is provided, and the storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the foregoing target person data processing method based on collected data.

[0010] In a third aspect of the present application, an electronic device is provided, which comprises a processor and the foregoing non-transitory computer readable storage medium.

[0011] The present application has at least the following beneficial effects:

[0012] The target personnel data processing method based on collected data provided in the application first generates a physical feature vector HT by collecting the physical data of the target personnel, and matches in the standard physical feature vector library, which can accurately locate the target standard physical feature vector MT that best matches the current physical feature of the personnel. After determining MT, the corresponding keyword list is used, which is based on the standard sentence library and accurately describes the causes of the standard physical feature. When obtaining the key video in combination with the standard behavior description sentence of the target personnel, this standardized cause description can accurately dock with the behavior description sentence obtained by processing the image through a large language model. The large language model classifies the scene type of the image collected in the target area to obtain a key image, and then obtains the video before and after the key image, ensuring that the selected key video contains the behavior scene closely related to the cause of the physical feature. From the coherence of data processing, the image acquisition device collects continuous video while collecting images, and intercepts the video before and after the key image as the key video based on the key image collection time, ensuring the time consistency of the video data, physical data and image data. The accurate correspondence in the time dimension makes the key video can completely present the behavior process that leads to a specific physical feature, providing complete and accurate materials for subsequent behavior analysis of the target personnel. Compared with the analysis method that only relies on a single data source or lacks accurate time association between data, the application greatly improves the reliability and comprehensiveness of behavior analysis through the close combination of multi-source data in time and content.

[0013] The final key video, because its content accurately corresponds to the behavior related to the cause of the physical feature of the personnel, can be used for targeted behavior analysis and reminding of the target personnel in actual application. For example, in the health management scene, if the key video shows that the personnel has the bad behavior of looking at the mobile phone for a long time, the system can push a reminding information in time based on this, inform the personnel of the health problems such as neck discomfort that may be caused by the behavior, and provide corresponding improvement suggestions. This reminding based on accurate data and analysis can attract more attention of the personnel than the non-targeted health tips, and promote them to change the bad behavior. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0015] Figure 1 The flowchart of the target personnel data processing method based on collected data provided in the embodiment of the application. DETAILED DESCRIPTION

[0016] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0017] It should be noted that the terms "first", "second" and the like in the description and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in other than the order illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0018] It should be noted that various aspects of the embodiments described below are within the scope of the appended claims. It should be apparent that the aspects described herein can be embodied in a wide variety of forms and that any specific structure and / or function described herein is merely illustrative. Based on the present application, a person of ordinary skill in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, an apparatus can be implemented and / or a method can be practiced using any number of the aspects set forth herein. In addition, such an apparatus and / or method can be implemented and / or practiced using other structure and / or functionality in addition to or other than one or more of the aspects set forth herein.

[0019] Please refer to Figure 1 The embodiments of the present application provide a target personnel data processing method based on collected data, the method comprises:

[0020] Step S100, collecting the physical data of the target personnel to obtain a physical feature vector HT.

[0021] Specifically, the physical data can be obtained through physical examination and the like. Here, after data cleaning, data standardization, feature selection and data coding and the like are performed on the physical data, a physical feature vector HT is constructed.

[0022] In step S200, the target keyword list is obtained according to the HT; wherein the target keyword list is a keyword list corresponding to the target standard sign feature vector MT determined from the HT among a plurality of standard sign feature vectors; the matching degree P between MT and HT MH is greater than the matching degree between any standard sign feature vector other than MT and HT, and P MH is greater than a preset matching degree threshold; the keywords in the keyword list are used to describe the causes of the standard sign corresponding to the standard sign feature vector; and the keywords are any standard sentence in a standard sentence library.

[0023] Specifically, if the matching degree between HT and any standard sign feature vector is greater than the matching degree between HT and other standard sign feature vectors, and the matching degree is greater than a preset matching degree threshold, it indicates that the situation of the target person is most consistent with the situation corresponding to the standard sign vector; each standard sign feature vector has a corresponding keyword list, and the keyword list corresponding to the target standard sign feature vector MT is determined as the target keyword list; here, the keywords in the keyword list are used to describe the causes of the standard sign corresponding to the standard sign feature vector. As an example, the keywords in the keyword list can be reading in a prone position, reading in a lying position, sitting crookedly, and hunchback, etc.; and the standard signs can be myopia, scoliosis, and other special situations. That is, reading in a prone position, reading in a lying position, etc. can be the cause of myopia.

[0024] In step S300, a plurality of key videos are obtained according to the target keyword list and the standard behavior description sentence corresponding to the target person; wherein the behavior description sentence is obtained by standardizing the output of a large language model according to the key images collected by the image acquisition device in the target area where the target person is located; the image acquisition device collects continuous videos at the same angle while collecting images; each key image has a corresponding collection time; the key video is a video corresponding to a preset time length before and after the collection time of the key image collected by the same image acquisition device; the key image is obtained by classifying the scene types of all images collected by the image acquisition device in the target time window according to the large language model; the key video is used for analyzing and reminding the behavior of the target person; and the time period corresponding to each key video is within the target time window.

[0025] Specifically, the standard behavior description statement corresponding to the target person is a statement describing the behavior characteristics of the target person. Here, the behavior description statement is obtained by standardizing the output of a large language model based on the key images collected by the image acquisition device in the target area where the target person is located. In this embodiment, as an example, the target area is a kindergarten, and the image acquisition device is a camera installed at each position in the kindergarten. Here, the working mechanism of the image acquisition device is to continuously record videos, but the recorded videos have a certain storage time. As an example, the storage time is 3 months, i.e., the target time window is 3 months. While recording videos in the target time window, the image acquisition device saves a picture every certain time interval. The key image is obtained by classifying the scene types of all images collected by the image acquisition device in the target time window based on a large language model, i.e., for the target keyword list, the behavior of the target person in some scenes is important, and the behavior in some scenes can be ignored. Therefore, by classifying the scene types of all images collected by the image acquisition device in the target time window based on a large language model, the key image is obtained, and then the key video is obtained based on the key image, thereby reducing the data processing amount.

[0026] As for the acquisition of the key video, the video corresponding to a preset time length before and after the collection time of the key image is intercepted. As an example, the time corresponding to a certain key image collected by device A is April 26, 2025, 2:00 pm; then, the key video can be the video collected by device A from 1:55 pm to 2:05 pm on April 26, 2025. Here, by analyzing the static key image, and then analyzing the behavior based on the key image to intercept the key video, the data processing amount is reduced, and the behavior analysis result obtained ultimately is more comprehensive and more accurate. The above-mentioned large model can be a GPT model, a BERT model, etc.

[0027] It can be understood that the key video is used to analyze and remind the behavior of the target person, i.e., the key video can contain some bad habits that cause the target standard body to remind the target person to correct these bad habits.

[0028] The target personnel data processing method based on collected data provided in the application first generates a physical feature vector HT by collecting physical feature data of a target personnel, and performs matching in a standard physical feature vector library, so as to accurately locate a target standard physical feature vector MT that best matches the current physical feature of the personnel. After determining the MT, the corresponding keyword list thereof is used, and these keywords are based on a standard sentence library and accurately describe the causes of the standard physical feature. When a key video is obtained in combination with a standard behavior description sentence of the target personnel, the standardized cause description can be accurately connected with the behavior description sentence obtained by processing the image by a large language model. The large language model classifies the scene type of the image collected in the target area to obtain a key image, and then obtains the video before and after the key image, so as to ensure that the selected key video contains the behavior scene closely related to the cause of the physical feature of the personnel. From the continuity of data processing, the image acquisition device collects continuous video while collecting the image, and intercepts the video before and after the key image as the key video based on the key image collection time, so as to ensure the time consistency of the video data, the physical feature data and the image data. The accurate correspondence in the time dimension enables the key video to completely present the behavior process that causes a specific physical feature to appear, and provides complete and accurate materials for subsequent behavior analysis of the target personnel. Compared with the analysis method that only relies on a single data source or lacks accurate time association between data, the application greatly improves the reliability and comprehensiveness of behavior analysis by closely combining multi-source data in time and content.

[0029] The finally obtained key video, because its content accurately corresponds to the behavior related to the cause of the physical feature of the personnel, can be used for targeted behavior analysis and reminding of the target personnel in actual application. For example, in a health management scene, if the key video shows that the personnel has the bad behavior of looking at the mobile phone for a long time, the system can push a reminding information in time based on this, inform the personnel of the health problems such as neck discomfort that may be caused by the behavior, and provide corresponding improvement suggestions. Compared with the health prompt without targeting, the reminding based on accurate data and analysis can attract more attention of the personnel and promote them to change the bad behavior.

[0030] In an exemplary embodiment of the application, the behavior description sentence corresponding to the target personnel is obtained according to the following steps:

[0031] In step S310, images collected by each image acquisition device in the target area of the target personnel in a target time window are obtained and image screening is performed to obtain N basic images; each image has a corresponding scene type; the scene type is obtained by scene analysis of the image by a large language model.

[0032] Specifically, each standard sign feature vector has a corresponding influence time; here, the influence time represents the key time of forming the standard sign feature vector. As an example: some standard signs can be more sudden, so the data within a time period close to the current time can be more important than the data far from the current time, so the key time can be shorter, and some standard signs can be chronic, so all the data within the target time window can be important, i.e. the key time can be longer. The target time window includes a plurality of equal and continuous target time periods; the image screening includes the following steps:

[0033] In step S311, according to the influence time corresponding to the target feature sign vector and the preset influence time-data collection ratio mapping table, the data collection ratio corresponding to each target time period within the target time window is obtained; wherein the preset influence time-data collection ratio mapping table includes a plurality of influence times, i.e. a target time period collection ratio list corresponding to each influence time; the target time period collection ratio list includes the data collection ratio of each target time period included in the target time window.

[0034] Here, the target time window is divided into a plurality of equal and continuous target time periods, and according to the influence time corresponding to the target feature sign vector and the preset influence time-data collection ratio mapping table, the data collection ratio corresponding to each target time period within the target time window is obtained. Each influence time has a corresponding target time period collection ratio list in the mapping table. As an example, if the length of the target time window is 3 months, the target time window is divided into six equal target time periods according to the chronological order; if the influence time corresponding to the target feature sign vector is 2 months, then for the last four target time periods (the four target time periods closer to the current time), the extraction ratio corresponding thereto is set to 100%, and for the first two target time periods (the two target time periods farther away from the current time), the extraction ratio of the target time period farthest from the current time is 20%, and the other is 40%. By controlling the data collection ratio, more reasonable data can be obtained, the effectiveness of the data can be improved, and the data processing amount can be reduced.

[0035] In step S312, according to the data collection ratio corresponding to each target time period, a corresponding proportion of images is randomly collected in each target time period of the target time window to obtain N basic images.

[0036] Here, according to the data collection ratio corresponding to each target time period, a corresponding proportion of images is randomly collected in each target time period of the target time window to obtain N basic images.

[0037] Step S320, clustering the N base images according to the scene types to obtain a base scene clustering cluster list JL=(JL1, JL2, …, JL x , …, JL y ); x=1, 2, …, y; where y is the number of base scene clustering clusters; JL x is the cluster identifier of the xth base scene clustering cluster; each base scene clustering cluster has a corresponding scene type; any two base images in each base scene clustering cluster correspond to the same scene type; any two base scene clustering clusters correspond to different scene types.

[0038] Here, each image has a corresponding scene type, as an example: the above scene types can be classrooms, playgrounds, canteens, etc., and the scene types are obtained by analyzing each image using a large language model. Images belonging to the same scene are clustered into a class, resulting in several base scene clustering clusters.

[0039] Step S330, according to the target keyword list and JL, obtaining a target base scene clustering cluster list M=(M1, M2, …, M g , …, M h ); g=1, 2, …, h; where h is the number of target base scene clustering clusters; M g is the cluster identifier of the gth target base scene clustering cluster; each target base scene clustering cluster corresponds to a scene type that is the same as any one of the key scene types corresponding to MT; the key scene types corresponding to MT are obtained by analyzing the target keyword list using a large language model.

[0040] Here, at least one key scene type corresponding to MT is obtained by analyzing the target keyword list using a large language model. As an example: if the keywords are lying down to read and lying down to read, then the key scene type obtained by analyzing the target keyword list using a large language model can be a classroom. If the keywords are slouching and hunchback, then the key scene types obtained by analyzing the target keyword list using a large language model can be a classroom, a canteen, a playground, etc. The target base scene clustering cluster is determined according to the corresponding base scene type of the obtained several clustering clusters, and the target base scene clustering cluster is the clustering cluster corresponding to the scene type with the highest relevance to the target person and the most likely to appear the behavior related to the target sign. Thus, further filtering of data is achieved, reducing the amount of data processing and improving data effectiveness.

[0041] Step S340, performing person analysis on each base image contained in M to obtain several intermediate base images; wherein the intermediate base image is a base image containing the target person.

[0042] Specifically, after determining the target basic scene clustering cluster, a person analysis is performed on all basic images contained in the clustering cluster, and the purpose of the task analysis is to screen out basic images containing target personnel. Further reduce the data processing amount and improve the data effectiveness.

[0043] In step S350, the large language model is used to perform standardized output on each intermediate basic image to obtain at least one standard behavior description sentence corresponding to each intermediate basic image.

[0044] Here, when the large language model is used to perform standardized output on each intermediate image, the input data of the large language model includes: the intermediate image, the output instruction, and the output word library; wherein the output instruction is used to describe the operation and output format for the intermediate basic image; and the output word library is used to provide output words for the output data.

[0045] Based on the above standardized output method, the obtained standard behavior description sentence is obtained according to the words in the output word library. As an example, the words contained in the output word library can be reading, lying, and sitting. The output instruction can be "select words from the output word library to describe the intermediate basic image, and output in a fixed output format". The fixed output format can be an adverbial part and a predicate part.

[0046] In an example embodiment of the present application, the standard behavior description sentence is any standard sentence in the standard sentence library; and according to the target keyword list and the standard behavior description sentence corresponding to the target personnel, a plurality of key videos are obtained, including:

[0047] In step S301, according to the target keyword list and each standard behavior description sentence, a plurality of key images are obtained; wherein the key image is an intermediate image corresponding to any standard behavior description sentence and any keyword in the target keyword list;

[0048] In step S302, according to the plurality of key images, a video corresponding to the same image acquisition device within a preset time length before and after the acquisition time of each key image is obtained to obtain a plurality of key videos.

[0049] In this embodiment, the key image is determined according to the standard behavior description sentence, and the key video is finally determined.

[0050] In an example embodiment of the present application, the behavior description sentence corresponding to the target personnel is obtained according to the following steps:

[0051] In step S360, a set of image lists C=(C1, C2, …, C i , …, C n); i = 1, 2, …, n; where n is the number of image acquisition devices in the target area where the target person is located; C i is the image list obtained by the i-th image acquisition device in the target area where the target person is located within the target time window; C i i，1 i，2 i，j i，m ); j = 1, 2, …, m; m is the number of images collected by the image acquisition device within the target time window; C i，j is the image identifier of the j-th image collected by the i-th image acquisition device in the target area where the target person is located within the target time window; the acquisition time interval between any two adjacent images corresponding to the same image acquisition device is the same.

[0052] Specifically, in the dimension of image acquisition device, all images collected by all image acquisition devices within the target time window are obtained.

[0053] Step S370, clustering all images in C according to the scene type to obtain a scene cluster list set AL = (AL1, AL2, …, AL c d ); c = 1, 2, …, d; where d is the number of scene clusters; AL c is the image list set corresponding to the c-th scene cluster; each scene cluster has a corresponding scene type; each scene type has a corresponding movement attribute; the movement attribute includes static attribute and dynamic attribute; any two images in each scene cluster correspond to the same scene type; any two scene clusters correspond to different scene types; AL c c，1 c，2 c，a c，k(c) ); a = 1, 2, …, k(c); k(c) is the number of image acquisition devices corresponding to the c-th scene cluster; k(c) < n; AL c，a is the image list collected by the a-th image acquisition device in the target time window in the scene type corresponding to the c-th scene cluster; AL c，a c，a,1 c，a，2 c，a，j c，a，m ); AL c，a，j is the image identifier of the j-th image collected by the a-th image acquisition device in the target time window in the scene type corresponding to the c-th scene cluster; the time of image acquisition by any two image acquisition devices is synchronous. ​​​​​​​​​​​​​

[0054] In this embodiment, all images are clustered, and the image list of each image acquisition device in the corresponding cluster is obtained in the dimension of the image acquisition device within the cluster. In this embodiment, each image acquisition device has a fixed shooting angle and does not rotate, and it corresponds to a unique scene type. For a scene type, there can be multiple corresponding image acquisition devices.

[0055] Step S380, if PAL c，a,b is equal to or greater than the preset image matching degree threshold, and the movement attribute of the scene type of the cth scene cluster is a static attribute, then all images corresponding to the a th image acquisition device in the scene type corresponding to the c th scene cluster are deleted or all images corresponding to the b th image acquisition device in the scene type corresponding to the c th scene cluster are deleted, to obtain AL’; PAL c，a,b is the comprehensive matching degree between the first preset number of images obtained by the a th image acquisition device in the scene type corresponding to the c th scene cluster and the first preset number of images obtained by the b th image acquisition device in the scene type corresponding to the c th scene cluster; b = 1, 2, …, k(c).

[0056] Here, PAL c，a,b is equal to or greater than the preset image matching degree threshold, and the movement attribute of the scene type of the c th scene cluster is a static attribute, that is, the similarity of the images obtained by two image acquisition devices in a certain scene type is very high, and the scene is a static attribute. Here, the static attribute means that the target person is less active in this scene type, and vice versa, the dynamic attribute means that the target person is more active in this scene and may walk around. If a certain scene type is a static attribute and the similarity of the images taken by two cameras is very high, then any one of them can be directly deleted. Data deduplication reduces data processing amount.

[0057] It can be understood that the above-mentioned comprehensive matching degree can be the average value of the similarity between the image (obtained by the a th image acquisition device) and the image (obtained by the b th image acquisition device) corresponding to each same time point in the preset number of images.

[0058] Step S390, according to the target keyword list and AL’, a target scene cluster list set D = (D1, D2, …, D r , …, D s ) is obtained; r = 1, 2, …, s; wherein s is the number of target scene clusters; D r is the image list set of the r th target scene cluster; the scene type corresponding to each target scene cluster is the same as at least one key scene type corresponding to MT; the key scene type corresponding to MT is obtained according to the target keyword list and the large language model.

[0059] Step S3100, performing person analysis on each image contained in D to obtain a plurality of intermediate images; wherein, the intermediate image is an image containing a target person in D.

[0060] Step S3110, performing standardized output on each intermediate image according to the large language model to obtain at least one standard behavior description sentence corresponding to each intermediate image.

[0061] Steps S3100-S3110 are the same as the principles of steps S340-350 described above, and will not be described here.

[0062] In an exemplary embodiment of the present application, after clustering all images in C according to the scene type to obtain a scene clustering cluster list set AL=(AL1, AL2, …, AL c , …, AL d ), the method further comprises:

[0063] Step S3120, if PAL c，a,b is less than the preset image matching degree threshold and the movement attribute of the scene type of the cth scene clustering cluster is a static attribute, then obtaining a target scene clustering cluster list D=(D1, D2, …, D r , …, D s ) according to the target keyword list and AL.

[0064] Specifically, if PAL c，a,b is less than the preset image matching degree threshold and the movement attribute of the scene type of the cth scene clustering cluster is a static attribute, it means that the two image acquisition devices belong to the same scene, but the images obtained are not similar, so both are retained, and the subsequent processing is the same as above, which will not be described here.

[0065] In an exemplary embodiment of the present application, after clustering all images in C according to the scene type to obtain a scene clustering cluster list set AL=(AL1, AL2, …, AL c , …, AL d ), the method further comprises:

[0066] Step S3130, if the movement attribute of the scene type of the cth scene clustering cluster is a dynamic attribute, then obtaining a target scene clustering cluster list D=(D1, D2, …, D r , …, D s ) according to the target keyword list and AL.

[0067] Specifically, if the movement attribute of the scene type of the cth scene cluster is a dynamic attribute, that is, the target person will perform more frequent activities in the scene, at this time, since the target person is relatively dynamic, the information obtained by each image acquisition device may change at different times, and therefore, all need to be retained. The subsequent processing is the same as above, and will not be repeated here.

[0068] In the example embodiments of the present application, an electronic device capable of implementing the above method is also provided.

[0069] Those skilled in the art can understand that various aspects of the present application can be implemented as a system, a method or a program product. Therefore, various aspects of the present application can be embodied as a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which can be collectively referred to as "circuitry", "module" or "system" here.

[0070] The electronic device according to this embodiment of the present application. The electronic device is only an example, and should not bring any limitation to the function and use range of the embodiments of the present application.

[0071] The electronic device is in the form of a general computing device. The components of the electronic device can include but are not limited to the above-mentioned at least one processor, the above-mentioned at least one storage, a bus connecting different system components including the storage and the processor.

[0072] The storage stores program code that can be executed by the processor, so that the processor executes the steps according to various example embodiments of the present application described in the above "example method" section of the present specification.

[0073] The storage can include a readable medium in the form of a volatile storage, such as a random access memory (RAM) and / or a cache storage, and can further include a read-only memory (ROM).

[0074] The storage can further include programs / utilities with a set of (at least one) program modules, such as an operating system, one or more application programs, other program modules, and program data, each of which or some combination of which can include the implementation of a network environment.

[0075] The bus can be one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of a variety of bus structures.

[0076] The electronic device can also be in communication with one or more external devices such as a keyboard or a pointing and / or selection device, and / or any devices (e.g., routers, modems, etc.) that enable a person to interact with the electronic device, and / or any devices (e.g., routers, modems, etc.) that enable the electronic device to communicate with one or more other computing devices. Such communication can occur via Input / Output (I / O) interface(s). Still yet, the electronic device can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or the Internet) through a network adapter. As illustrated, the network adapter is in communication with the other components of the electronic device through a bus. It should be appreciated that although not shown, other hardware and / or software modules could be used in connection with the electronic device. These include, but are not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc.

[0077] Those skilled in the art will readily understand that the example embodiments described herein can be implemented by software and / or by software in combination with the necessary hardware. Thus, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash disk, a mobile hard disk, etc.) or a network, and includes a number of instructions to enable a computing device (such as a personal computer, a server, a terminal device, or a network device, etc.) to perform the methods according to the embodiments of the present application.

[0078] In the example embodiments of the present application, a computer readable storage medium is also provided, which stores the program product capable of implementing the methods described above. In some possible embodiments, various aspects of the present application can also be implemented in the form of a program product, which includes program codes for causing a terminal device to perform the steps described in the “example method” section above according to various example embodiments of the present application when the program product is run on the terminal device.

[0079] The program product can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium may, for example, be but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0080] Computer readable signal media can include a propagated data signal with instructions embodied in data signals. Such propagated signal can take a wide variety of forms, including but not limited to electro-magnetic signals, optical signals, and so forth. Such computer readable storage media further includes any tangible storage media that is not yet in use by a human being, which explicitly excludes propagated signals per se.

[0081] Program code embodied on a computer readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0082] Program code, used by or in connection with the routines described herein, can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0083] In addition, the flow diagrams illustrate the functionality of the various embodiments of the present application. It is noted that the steps of the flow diagrams can be performed in a different order than that shown in the figures. In addition, some steps can be performed simultaneously. In addition, some steps can be performed by different modules or components than those shown in the figures.

[0084] It should be noted that, although the above detailed description refers to several modules or units of the device for action execution, such a division is not mandatory. Indeed, according to an embodiment of the application, the features and functionalities of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functionalities of one module or unit described above can be further divided into several modules or units embodied.

[0085] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any changes or replacements within the technical scope disclosed by the present application, which can be easily thought by any person skilled in the art, should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A target person data processing method based on collected data, characterized by, The method comprises: Collecting the vital sign data of the target person to obtain a vital sign feature vector HT; According to the HT, a target keyword list is acquired; wherein the target keyword list is a keyword list corresponding to a target standard sign feature vector MT determined by the HT from a plurality of standard sign feature vectors; a matching degree P between MT and HT MH is greater than a matching degree between any standard sign feature vector other than MT and HT, and P MH is greater than a preset matching degree threshold; the keywords in the keyword list are used to describe the causes of the standard sign corresponding to the standard sign feature vector; and the keywords are any standard sentence in a standard sentence library. According to the target keyword list and the standard behavior description sentence corresponding to the target person, a plurality of key videos are obtained; the behavior description sentence is obtained by standardizing the output of a large language model based on the key images collected by the image acquisition device in the target area where the target person is located; the image acquisition device collects continuous videos at the same angle while collecting images; each key image has a corresponding collection time; the key video is the video corresponding to the preset time length before and after the collection time of the key image collected by the same image acquisition device; the key image is obtained by classifying the scene type of all images collected by the image acquisition device in the target time window based on the large language model; the key video is used to analyze and remind the behavior of the target person; the time period corresponding to each key video is within the target time window.

2. The target person data processing method based on collected data according to claim 1, characterized by, The behavior description sentence corresponding to the target person is obtained according to the following steps: Collecting the images collected by each image acquisition device in the target area where the target person is located within the target time window and performing image screening to obtain N basic images; each image has a corresponding scene type; The scene type is obtained by performing scene analysis on the image based on the large language model; According to the scene type, the N basic images are clustered to obtain a basic scene clustering cluster list JL=(JL1, JL2, …, JL x , …, JL y ); x=1, 2, …, y; wherein y is the number of the basic scene clustering clusters; JL x is the cluster identifier of the xth basic scene clustering cluster; each basic scene clustering cluster has a corresponding scene type; the scene types corresponding to any two basic images in each basic scene clustering cluster are the same; the scene types corresponding to any two basic scene clustering clusters are different; Based on the target keyword list and JL, the target basic scene cluster list M = (M1, M2, ..., M...) is obtained. g M h ); g = 1, 2, ..., h; where h is the number of clusters in the target basic scene; M g This is the cluster identifier for the g-th target basic scene cluster; the scene type corresponding to each target basic scene cluster is the same as any key scene type corresponding to MT; the key scene type corresponding to MT is obtained based on the target keyword list and the large language model; Performing person analysis on each basic image contained in M to obtain a plurality of intermediate basic images; the intermediate basic image is a basic image containing the target person; Standardizing the output of each intermediate basic image based on the large language model to obtain at least one standard behavior description sentence corresponding to each intermediate basic image.

3. The target person data processing method based on collected data according to claim 2, characterized by, When the large language model is used to standardize the output of each intermediate image, the input data of the large language model includes: the intermediate image, the output instruction and the output word library; the output instruction is used to describe the operation and output format performed on the intermediate image; the output word library is used to provide output words for the output data.

4. The target person data processing method based on collected data according to claim 3, characterized by, The standard behavior description sentence is any standard sentence in the standard sentence library; according to the target keyword list and the standard behavior description sentence corresponding to the target person, a plurality of key videos are obtained, including: According to the target keyword list and each standard behavior description sentence, a plurality of key images are obtained; the key image is an intermediate image corresponding to any standard behavior description sentence and any keyword in the target keyword list; According to the plurality of key images, the video corresponding to the preset time length before and after the collection time of each key image collected by the same image acquisition device is obtained to obtain a plurality of key videos.

5. The target person data processing method based on collected data according to claim 3, characterized by, Each standard vital sign feature vector has a corresponding influence time; the target time window includes a plurality of equal and continuous target time periods; the image screening includes the following steps: According to the influence time corresponding to the target feature sign vector and the preset influence time-data collection ratio mapping table, a data collection ratio corresponding to each target time period in the target time window is obtained; wherein the preset influence time-data collection ratio mapping table contains a plurality of influence times, that is, a target time period collection ratio list corresponding to each influence time; the target time period collection ratio list contains a data collection ratio of each target time period included in the target time window; According to the data collection ratio corresponding to each target time period, a corresponding proportion of images is randomly collected in each target time period of the target time window to obtain N basic images.

6. The target person data processing method based on collected data according to claim 1, characterized by, The behavior description sentence corresponding to the target person is obtained according to the following steps: Obtain a list set C = (C1, C2, ..., Cn) of images acquired by each image acquisition device within the target time window within the target area where the target person is located. i C n ); i = 1, 2, ..., n; where n is the number of image acquisition devices within the target area where the target person is located; C i C is a list of images obtained from the i-th image acquisition device within the target area where the target person is located, acquired within the target time window; i =(C i，1 C i，2 C i，j C i，m ); j = 1, 2, ..., m; m is the number of images acquired by the image acquisition device within the target time window; C i，j The image identifier is the j-th image captured by the i-th image acquisition device within the target time window in the target area where the target person is located; the acquisition time interval is the same between any two adjacent images acquired by the same image acquisition device. Cluster all images in C according to scene type to obtain a list set of scene clusters AL = (AL1, AL2, ..., AL3). c , ..., AL d ); c = 1, 2, ..., d; where d is the number of scene clusters; AL c This is the image list set corresponding to the c-th scene cluster; each scene cluster has a corresponding scene type; each scene type has a corresponding motion attribute; the motion attribute includes static and dynamic attributes; any two images within a scene cluster correspond to the same scene type; any two scene clusters correspond to different scene types; AL c =(AL c，1 AL c，2 , ..., AL c，a , ..., AL c，k(c) ); a = 1, 2, ..., k(c); k(c) is the number of image acquisition devices corresponding to the c-th scene cluster; k(c) < n; AL c，a This is the list of images acquired by the image acquisition device of the a-th image acquisition device within the target time window, corresponding to the c-th scene cluster; AL c，a =(AL c，a,1 AL c，a，2 , ..., AL c，a，j , ..., AL c，a，m ); AL c，a，j This is the image identifier of the j-th image acquired by the a-th image acquisition device within the target time window in the scene type corresponding to the c-th scene cluster; the acquisition times of any two image acquisition devices are synchronized; if PAL c，a,b is equal to or greater than a preset image matching degree threshold value, and a movement attribute of a scene type of the cth scene clustering cluster is a static attribute, then all images corresponding to the a th image acquisition device in the scene type corresponding to the cth scene clustering cluster or all images corresponding to the b th image acquisition device in the scene type corresponding to the cth scene clustering cluster are deleted, to obtain AL’; PAL c，a,b is a comprehensive matching degree between the first preset number of images acquired by the a th image acquisition device in the scene type corresponding to the cth scene clustering cluster and the first preset number of images acquired by the b th image acquisition device in the scene type corresponding to the cth scene clustering cluster; b = 1, 2, …, k(c). Based on the target keyword list and AL', the target scene cluster list set D = (D1, D2, ..., D...) is obtained. r D s ); r = 1, 2, ..., s; where s is the number of clusters in the target scene; D r Let r be the image list set for the r-th target scene cluster; the scene type corresponding to each target scene cluster is the same as at least one key scene type corresponding to MT; the key scene type corresponding to MT is obtained based on the target keyword list and the large language model; Performing character analysis on each image contained in D to obtain a plurality of intermediate images; wherein the intermediate image is an image containing the target person in D; According to the large language model, a standardized output is performed on each intermediate image to obtain at least one standard behavior description sentence corresponding to each intermediate image.

7. The target person data processing method based on collected data according to claim 6, characterized by, After clustering all images in C according to the scene type to obtain a list of scene clustering clusters AL = (AL1, AL2,..., AL c ,..., AL d ), the method further comprises: If PAL c，a,b If the scene type of the cth scene cluster is less than the preset image matching degree threshold and the moving attribute of the scene type is a static attribute, a target scene cluster list D=(D1, D2, …, D r , …, D s ) is obtained according to the target keyword list and the AL.

8. The target person data processing method based on collected data according to claim 6, characterized by, Cluster all images in C according to scene type to obtain a list set of scene clusters AL = (AL1, AL2, ..., AL3). c , ..., AL d Following this, the method further includes: If the scene type mobility attribute of the c-th scene cluster is a dynamic attribute, then based on the target keyword list and AL, the target scene cluster list D = (D1, D2, ..., Dc) is obtained. r D s ). 9.A non-transitory computer readable storage medium having stored therein at least one instruction, characterized in that, The at least one instruction is loaded and executed by the processor to implement the method of any one of claims 1-8.

10. An electronic device, comprising: The non-transitory computer readable storage medium described in claim 9 is included. The non-transitory computer readable storage medium described in claim 9 is included.

Images