Camera intelligent algorithm update methods, electronic devices, and storage media

Abstract

This disclosure provides a method, electronic device, and storage medium for updating a camera's intelligent algorithm. The method includes: acquiring scene type information of the camera in a video surveillance system; sending a scene-type-related intelligent algorithm to the camera based on the scene type information, enabling the camera to perform business analysis based on the received intelligent algorithm; and triggering and updating the intelligent algorithm based on the analysis results data fed back by the camera based on the intelligent algorithm during the use of the video surveillance system. This achieves the goal of automatically updating and iterating the camera's intelligent algorithm during the use of the surveillance system.

Description

Technical Field

[0001] This disclosure relates to the field of video surveillance technology, and in particular to a method for updating a camera's intelligent algorithm, an electronic device, and a storage medium. Background Technology

[0002] Current intelligent algorithms for IP cameras (network cameras) often require collecting massive amounts of footage, calibrating and training it, and then optimizing it to develop a universally applicable algorithm. The IPC then uses this algorithm to perform intelligent recognition and provides information to the video surveillance system for business applications. When misidentification occurs, it's often necessary to collect the misidentified footage again, train and optimize the algorithm, and then provide it to the IPC for upgrades. This entire cycle is lengthy, data sharing is difficult, and optimization for one scenario may lead to problems in other situations. Summary of the Invention

[0003] This disclosure provides a method, electronic device, and storage medium for updating the intelligent algorithm of a camera, so as to automatically update and iterate the intelligent algorithm of the camera during the use of a video surveillance system.

[0004] In a first aspect, embodiments of this disclosure provide a method for updating a camera's intelligent algorithm, comprising:

[0005] Obtain scene type information of the camera in the video surveillance system;

[0006] Based on the scene type information of the camera, a scene-type related intelligent algorithm is sent to the camera, enabling the camera to perform business analysis based on the received intelligent algorithm;

[0007] During the use of the video surveillance system, the intelligent algorithm is triggered and updated based on the analysis results data fed back by the camera based on the intelligent algorithm.

[0008] Secondly, embodiments of this disclosure also provide an electronic device, including:

[0009] One or more processors;

[0010] Memory, used to store one or more programs.

[0011] When one or more programs are executed by one or more processors, the one or more processors implement a camera intelligent algorithm update method as described in any embodiment of this disclosure.

[0012] Thirdly, embodiments of this disclosure also provide a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements a method for updating a camera intelligent algorithm as described in any embodiment of this disclosure.

[0013] In this embodiment of the disclosure, the corresponding intelligent algorithm is first sent to the camera according to the scene in which the camera is located in the monitoring system. Then, during the use of the monitoring system, the corresponding intelligent algorithm is triggered and updated according to the historical analysis results fed back by the camera based on the intelligent algorithm. This achieves the purpose of automatically updating and iterating the camera's intelligent algorithm during the use of the monitoring system. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the update method of the camera intelligent algorithm in the first embodiment of this disclosure;

[0015] Figure 2 This is a flowchart illustrating a method for updating a camera intelligent algorithm in a second embodiment of this disclosure;

[0016] Figure 3 This is a flowchart illustrating an update method for another camera intelligent algorithm in the third embodiment of this disclosure;

[0017] Figure 4 This is a schematic diagram of the structure of the camera intelligent algorithm update device in the fourth embodiment of this disclosure;

[0018] Figure 5 This is a schematic diagram of the structure of an electronic device that implements the update method of the camera intelligent algorithm in the fifth embodiment of this disclosure. Detailed Implementation

[0019] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the drawings, not all structures.

[0020] Figure 1 This is a flowchart of a method for updating a camera intelligent algorithm according to the first embodiment of the present disclosure. This embodiment is applicable to the situation of automatically updating the intelligent algorithms used by each camera in a monitoring system. The method can be executed by a camera intelligent algorithm updating device, which can be implemented in software and / or hardware and can be integrated into an electronic device, such as a training server within a monitoring system.

[0021] like Figure 1 As shown, the update method for the camera's intelligent algorithm specifically includes the following process:

[0022] S101. Obtain the scene type information of the camera in the video surveillance system.

[0023] S102. Based on the scene type information of the camera, send a scene-type related intelligent algorithm to the camera, so that the camera can perform business analysis based on the received intelligent algorithm.

[0024] In this embodiment of the disclosure, the scene type information includes the camera's geographical location information (including intersections, inside stations, station entrances and exits, community entrances and exits, indoor halls, indoor corridors, etc.), the camera's service type (face, human / vehicle, behavior analysis, etc.), and resolution, etc.

[0025] Upon obtaining the scene type information of the camera, an intelligent algorithm related to the scene type is sent to the camera, enabling the camera to perform business analysis based on the received intelligent algorithm. For example, if the scene type information of a certain camera is face recognition business, a face recognition algorithm is sent to that camera, allowing the camera to analyze the face data collected by the camera according to the face recognition algorithm.

[0026] S103. During the use of the video surveillance system, the intelligent algorithm is triggered and updated based on the analysis results data fed back by the camera based on the intelligent algorithm.

[0027] In this embodiment, since the analysis results obtained by the camera based on the intelligent algorithm may be inaccurate, the analysis result data fed back by the camera based on the intelligent algorithm includes both accurately identified result data and incorrectly identified result data. The incorrectly identified result is determined based on feedback from the monitoring system user. It should be noted that after receiving feedback on the incorrect identification, the monitoring system administrator will correct the incorrect identification result, and a result correction record will be obtained after each correction.

[0028] In this embodiment, if a camera reports a large number of incorrect identification results (i.e., a large number of result correction records), it indicates that the intelligent algorithm within the camera is not suitable for the scene in which the camera is located, and the monitoring system needs to update and iterate the intelligent algorithm within the camera. Therefore, in an optional implementation, the number of result correction records is used as a marker to trigger an intelligent algorithm update. During the use of the video surveillance system, if the video management server in the video surveillance system detects that the number of result correction records included in the analysis result data reported by a certain camera based on the target intelligent algorithm reaches a threshold, it sends an update instruction for the target intelligent algorithm to the training server in the video surveillance system. This causes the training server to trigger and update the target intelligent algorithm according to the update instruction, and the data in the result correction records can be directly used as training samples during training.

[0029] In this embodiment of the disclosure, the corresponding intelligent algorithm is first sent to the camera according to the scene in which the camera is located in the monitoring system. Then, during the use of the monitoring system, the corresponding intelligent algorithm is triggered and updated according to the historical analysis results fed back by the camera based on the intelligent algorithm. This achieves the purpose of automatically updating and iterating the camera's intelligent algorithm during the use of the monitoring system.

[0030] Figure 2 This is a flowchart of a camera intelligent algorithm update method according to a second embodiment of the present disclosure, see [link to flowchart]. Figure 2 The method includes:

[0031] S201. During the use of the video surveillance system, receive update instructions for the target intelligent algorithm issued by the video management server.

[0032] In this embodiment, the video surveillance system includes a video management server, at least one camera (IPC) deployed with intelligent algorithms, non-intelligent cameras, and a GPU server. The GPU server's primary function is to perform streaming analysis on cameras that do not support intelligent features. It also reserves some GPU servers for training and optimizing algorithms for existing intelligent cameras. In other words, this embodiment uses a portion of the reserved GPU servers in the surveillance system as training servers to ensure that the intelligent algorithms in the intelligent cameras can be automatically optimized during the use of the surveillance system.

[0033] Furthermore, in the video surveillance system, intelligent cameras provide analysis results during operation and store these results in the system's big data. Non-intelligent cameras are streamed to a backend GPU server for intelligent analysis, and the results are also recorded in the system's big data. To better record the analysis results, this embodiment provides an answer record table (i.e., Table_Record_Answer table) for recording these results. Each record in the answer record table includes at least the camera identifier (e.g., camera ID), time point, algorithm identifier (algorithm ID), and analysis result (e.g., the target type and target coordinates analyzed by the camera). See Table 1 for an example, which shows the specific content recorded in the answer record table.

[0034] Table 1. Answer Record Sheet

[0035]

[0036]

[0037] According to the first record in the answer record table, the camera identified as CamID_1 captured a picture of a male wearing glasses at 08:10:05 on 2021-05-23.

[0038] It should be noted that, due to the possibility of inaccurate analysis results obtained by cameras based on intelligent algorithms, the answer record table needs to be gradually improved during the use of the monitoring system, that is, the analysis results need to be corrected. Therefore, the answer record table should also include information on result correction (such as correction category, correction content, and the person making the correction). In specific implementation, a system correction permission is added to the video surveillance system. Users with this permission can add fully structured information to the videos or images they access during system use. These users can re-mark the screen when using recordings or images, including adding targets, deleting targets, modifying target attributes, and modifying target coordinates. For example, see Table 2, which shows the answer record table with result correction information.

[0039] Table 2. Answer Record Sheet

[0040]

[0041] As shown in Table 2, the first record in the answer record table corrected the target attribute, and the second record corrected the target coordinate. Therefore, both the first and second records in the answer record table are result correction records.

[0042] In this embodiment of the disclosure, if the number of result correction records for a target intelligent algorithm (e.g., Algorithm A) reaches a preset threshold within a certain period, it indicates that the accuracy of the analysis results obtained based on the target intelligent algorithm is poor, and the target intelligent algorithm needs to be iteratively updated. At this time, the video management server in the video surveillance system can send an update command for the target intelligent algorithm to the training server. That is, the update command is issued after the number of result correction records included in the analysis results based on the target intelligent algorithm reaches the threshold.

[0043] S202. According to the update instruction, obtain the camera identifier and time point included in each result correction record.

[0044] In one optional implementation, after receiving an update instruction, the training server retrieves the camera identifier and time point included in each result correction record from the answer record table used to store the analysis results of the monitoring system. The purpose is to determine the scene type and tag information of the recording at that time point for each result correction record based on the camera identifier and time point. Here, a result correction record refers to a corrected analysis result record in the answer record table, and each result correction record includes at least the camera identifier, time point, intelligent algorithm identifier, analysis result information, and correction result information. For example, the specific content of the result correction record can be seen in the first record of Table 2.

[0045] S203. Based on the camera identifier and time point included in each result correction record, determine the scene type information and video recording tag information of the camera corresponding to each result correction record.

[0046] In this embodiment of the disclosure, the video surveillance system includes a camera scene type table (Table_Scene_Info) for recording the scene in which each camera is located. Each record in the camera scene type table includes at least the camera identifier, the identifier of the intelligent algorithm deployed within the camera, and the scene type information of the camera. The scene type information includes the camera's geographical location information (including intersections, inside stations, station entrances / exits, community entrances / exits, indoor halls, indoor corridors, etc.), the camera's service type (face, human / vehicle, behavior analysis, etc.), and resolution, etc.

[0047] In one optional implementation, field A1 identifies the camera's geographic location information. Therefore, an enumerated value of field A1 can represent a specific geographic location; for example, A1 with a value of 0 indicates an intersection, A1 with a value of 1 indicates within a station, and A1 with a value of 2 indicates a station entrance / exit. Field A2 identifies the camera's service type; for example, A2 with a value of 0 indicates face recognition, 1 indicates a vehicle / non-vehicle / human vehicle, and 2 indicates behavior analysis. Field A3 identifies the resolution (mainly including 200W / 300W / 800W / 4K, etc.); for example, a value of A3 of 0 indicates a resolution of 200W.

[0048] For example, if a camera identified as CamID_1 is a 200W resolution camera located at the station entrance / exit primarily capturing human faces, then the corresponding record in the camera scene type table (Table_Scene_Info) would be as follows:

[0049] Camera ID A1 A2 A3 Current algorithm CamID_1 2 0 0 Algorithm A

[0050] Based on the above, after obtaining the camera identifiers included in each result correction record, the camera scene type information corresponding to each correction record can be determined by querying the camera scene type table (e.g., by matching camera identifiers) according to the camera identifiers included in each result correction record.

[0051] Furthermore, in this embodiment of the disclosure, the video surveillance system also includes a recording category table (Table_Record_Info), primarily used to record the tag information of video recordings in the surveillance system. The recording category table is updated based on data from the answer record table (Table_Record_Answer), and each record in the recording category table includes at least the camera identifier, time range (Time_Range), and video recording tag information. The video recording tag information includes a subdivided target type (Obj_Type) and a natural type (Nature_type).

[0052] For example, the structure of the video recording category table is as follows, with one record marked for each camera every half hour.

[0053] Camera ID Time_Range Obj_Type Nature_type CamID_1 08:00:00-08:29:59 B1B2B3B4 C1C2

[0054] The `Obj_Type` field is a further refinement of field A2 in the `Table_scene_info` table, using B1, B2, B3, and B4. For example, B1 identifies motor vehicles, B2 identifies non-motor vehicles, and B3 identifies people. Specifically, B1: 0 indicates no motor vehicles, 1 indicates cars, 2 indicates trucks, and 3 indicates both cars and trucks, etc.; B2: 0 indicates no non-motor vehicles, 1 indicates non-motor vehicles, and its enumeration values ​​are expandable; B3: 0 indicates no people, 1 indicates people wearing glasses, 2 indicates people wearing hats, etc., 3 indicates faces, and 4 indicates people with multiple features, etc., and its enumeration values ​​are expandable; B4: identifies the main category of the material, for example, 0 indicates a large proportion of motor vehicles, and 1 indicates a large proportion of people.

[0055] In the Nature_type field, the C1 bit enumeration value is 0 for daytime, 1 for nighttime, and 2 for alternating day and night; the C2 bit is 0 for normal, 1 for rainy day, 2 for backlight, and 3 for infrared, etc.

[0056] It should be noted that in this embodiment, the video recording tags are updated according to the Table_Record_Answer in the current video surveillance system to tag the recordings of the current system, forming the Table_Record_Info table. For example, a camera identified as CamID_1 is a 200W resolution camera deployed at the entrance / exit of a station, mainly for recognizing faces and bodies. If this camera detected a large number of people during the 08:00:00-08:29:59 segment of the video surveillance system, then the record for this segment would be:

[0057] Camera ID Time_Range Obj_Type Nature_type CamID_1 08:00:00-08:29:59 0031 00

[0058] Based on the video recording category table (Table_Record_Info), the time points included in each result record can be corrected. By querying the video recording category table, the time range in which the time points fall can be determined, and then the corresponding video recording tag information for each corrected record can be determined.

[0059] For example, a camera identified as CamID_1 has a result correction record at time 08:10:05 on 2021-05-24. The camera scene type information of CamID_1 in the camera scene type table (Table_Scene_Info) is A1A2A3, and the field value is 200. The time point of the result correction record corresponds to the Obj_Type and Nature_type in the video recording tag information of 08:00:00-08:29:59 in the recording category table (Table_Record_Info) as 0031 and 00, respectively.

[0060] S204. Based on each result, correct the scene type information and video recording tag information of the corresponding camera, obtain video training materials from the monitoring system, and retrain the target intelligent algorithm using the video training materials to obtain a new intelligent algorithm.

[0061] In one optional implementation, for any result, the video recording tag information (e.g., Obj_Type and Nature_type information) corresponding to the record is corrected, and video recordings with the same video recording tag information are determined as the first type of video training material according to the video recording category table. That is, video recordings with the same Obj_Type and Nature_type are found as the first type of video training material.

[0062] For any given result, the scene type information corresponding to the record is corrected. Based on the camera scene type table, a target camera with the same scene type information is identified. Then, based on the target camera's representation, the video recording captured by that target camera is determined in the recording type table and used as the second type of video training material. Subsequently, the target intelligent algorithm is retrained using both the first and second types of video training material to obtain a new intelligent algorithm. It should be noted that the reason for choosing video recording materials with the same labels and the same scenes to retrain the target intelligent algorithm is to improve the accuracy of the trained intelligent algorithm.

[0063] Furthermore, after obtaining the new intelligent algorithm, it is necessary to verify the recognition accuracy of the new intelligent algorithm to determine whether the new intelligent algorithm meets the conditions. In an optional implementation, the new intelligent algorithm can be used to analyze the newly generated video recordings and determine whether the accuracy of the analysis results meets the preset conditions, such as whether the accuracy is greater than a preset value; if so, S105 is executed; otherwise, the above steps are returned to continue training.

[0064] S205. Notify the video management server to update the new intelligent algorithm to all cameras using the target intelligent algorithm.

[0065] Optionally, the video management server can determine which cameras are using the target intelligent algorithm based on the camera scene type table (Table_Scene_Info table), and then deploy the new intelligent algorithm on these cameras. In other words, the new intelligent algorithm replaces the target intelligent algorithm, so that these cameras can provide feedback analysis results based on the new intelligent algorithm.

[0066] In this embodiment, the intelligent algorithm of the camera in the current system is updated and iterated based on the difference information during the use of the video surveillance system. Moreover, the material required for the intelligent algorithm update and iteration is obtained from the monitoring system based on the camera scene type information and video recording tag information corresponding to each result correction record. This realizes the calibration of training material based on the historical analysis results of the monitoring system itself, and then the target intelligent algorithm is retrained using the training material. Thus, the purpose of automatically updating and iterating the intelligent algorithm of the camera during the use of the monitoring system is achieved.

[0067] Figure 3 This is a flowchart of a camera intelligent algorithm update method according to a third embodiment of the present disclosure, see [link to flowchart]. Figure 3 The method includes:

[0068] S301. For any newly added camera in the video surveillance system, obtain the scene type information of the newly added camera.

[0069] S302. Determine the intelligent algorithm used by other cameras that have the same camera scene type information as the newly added camera from the camera scene type table.

[0070] When adding a camera to a video surveillance system, it's necessary to select the optimal intelligent algorithm for it. To choose the optimal algorithm, it's essential to understand the scene type of the new camera. This scene type is determined during the camera's deployment. After obtaining the scene type, the system then uses intelligent algorithms from the camera scene type table (Table_Scene_Info) to identify other cameras with the same scene type information as the new camera.

[0071] S303. Select an intelligent algorithm from the intelligent algorithms used by other cameras and deploy it to the new camera.

[0072] In this embodiment of the disclosure, if other cameras use the same type of intelligent algorithm, that is, other cameras with the same scene type use the same intelligent algorithm, then it is only necessary to deploy the intelligent algorithm used by the other cameras to the newly added camera.

[0073] If other cameras use different types of intelligent algorithms (i.e., other cameras with the same scene type use at least two intelligent algorithms), then the intelligent algorithm that best suits the new camera needs to be selected from these at least two. In practice, this involves first statistically analyzing the reference analysis results of each intelligent algorithm within a preset time period. These reference analysis results are obtained by statistically analyzing the answer record table. Then, based on each intelligent algorithm, the video recordings captured by the new camera within the preset time period are analyzed to obtain at least two sampling analysis results. Finally, based on the similarity between the reference analysis results and sampling analysis results corresponding to each intelligent algorithm within the same time period—that is, for any given intelligent algorithm, the similarity between the reference analysis results obtained using that intelligent algorithm and the sampling analysis results—an intelligent algorithm is selected and deployed to the new camera. For example, the intelligent algorithm with the highest similarity value is selected and deployed to the new camera.

[0074] To illustrate this process in detail, let's assume that by comparing the camera scene type table (Table_Scene_Info), we determine that other cameras with the same camera scene type information as the newly added camera use three intelligent algorithms: X1, X2, and X3. Based on the answer record table (Table_record_answer), we determine the reference analysis results for each algorithm within a preset time period (which can be multiple time periods).

[0075] For example, regarding the intelligent algorithm X1:

[0076] During the cumulative period from 08:00:00 to 09:00:00 over 30 days, the analysis results of CamID1 were (face: 0.8; body: 0.1; motor vehicle: 0.05; non-motor vehicle: 0.05), and the values ​​represent the percentage of each data type.

[0077] The analysis results of CamID1 from 12:00:00 to 13:00:00 over a cumulative period of 30 days are as follows: (face: 0.85; body: 0.1; motor vehicle: 0.00; non-motor vehicle: 0.05), and the values ​​represent the proportion of each data type.

[0078] The analysis results of CamID1 over the cumulative period of 30 days from 20:00:00 to 21:00:00 are (face: 0.75; body: 0.2; motor vehicle: 0.05; non-motor vehicle: 0.05), with each value representing the percentage of that data type. Therefore, the reference analysis results of the intelligent algorithm X1 in the preset time periods (8-9 am, 12-1 pm, 8-9 pm) are as follows: (0.8, 0.1, 0.05, 0.05), (0.85, 0.1, 0, 0.05), and (0.75, 0.2, 0.05, 0.05).

[0079] The X2 and X3 algorithms are similar and will not be described in detail here. The resulting segmented statistical table (Table_Alg_Info) showing the reference analysis results for different intelligent algorithms is as follows:

[0080] algorithm 8-9 o'clock 12-13 o'clock 8-9 PM X1 (0.8,0.1,0.05,0.05) (0.85,0.1,0,0.05) (0.75,0.2,0.05.0.05) X2 () () () X3 () () ()

[0081] Furthermore, the training server sequentially calls intelligent algorithms X1, X2, and X3 to sample and analyze the recordings from the newly added cameras within a preset time period. For example, when intelligent algorithm X1 is called to analyze the recordings within the preset time period (e.g., 8-9 AM, 12-1 PM, 8-9 PM), the sampling and analysis results are as follows: (0.7, 0.2, 0.05, 0.05), (0.8, 0.1, 0.05, 0.05), and (0.8, 0.1, 0.05, 0.05).

[0082] For algorithm X1, when calculating the similarity between the reference analysis result and the sampling analysis result, the similarity between the reference analysis result (0.8, 0.1, 0.05, 0.05) and the sampling analysis result (0.7, 0.2, 0.05, 0.05) is calculated sequentially; the similarity between the reference analysis result (0.85, 0.1, 0, 0.05) and the sampling analysis result (0.8, 0.1, 0.05, 0.05) is calculated; and the similarity between the reference analysis result (0.75, 0.2, 0.05, 0.05) and the sampling analysis result (0.8, 0.1, 0.05, 0.05) is calculated. It should be noted that in this embodiment, the similarity calculation uses cosine similarity. Finally, the mean of the three calculated similarities is taken as the matching degree between intelligent algorithm X1 and the newly added camera. Similarly, the matching degree between intelligent algorithm X2 and the newly added camera, as well as the matching degree between intelligent algorithm X3 and the newly added camera, can be calculated. Finally, the intelligent algorithm with the highest matching degree is deployed to the newly added cameras.

[0083] In this embodiment of the disclosure, when a new camera is added to the monitoring system, the optimal intelligent algorithm can be selected for the new camera by calculating similarity.

[0084] Figure 4This is a schematic diagram of a camera intelligent algorithm updating device according to the fourth embodiment of this disclosure. This embodiment is applicable to situations where intelligent algorithms used by various cameras in a monitoring system are automatically updated. See also... Figure 4 The device includes:

[0085] The acquisition module is used to acquire scene type information of the camera in the video surveillance system;

[0086] The delivery module is used to send intelligent algorithms related to the scene type to the camera based on the scene type information of the camera, so that the camera can perform business analysis based on the received intelligent algorithms;

[0087] The update module is used to trigger and update the intelligent algorithm based on the analysis results data fed back by the camera based on the intelligent algorithm during the use of the video surveillance system. Optionally, based on the above embodiments, the update module is specifically used for:

[0088] During the use of the video surveillance system, if the video management server in the video surveillance system detects that the number of result correction records included in the analysis result data based on the target intelligent algorithm has reached a threshold, it will issue an update instruction for the target intelligent algorithm to the training server in the video surveillance system, so that the training server will trigger and update the target intelligent algorithm according to the update instruction.

[0089] Based on the above embodiments, optionally, the updating module includes:

[0090] The first data determination unit is used to obtain the camera identifier and time point included in each result correction record according to the update instruction;

[0091] The second data determination unit is used to determine the scene type information and video recording tag information of the camera corresponding to each result correction record based on the camera identifier and time point included in each result correction record.

[0092] The material acquisition and training unit is used to correct the scene type information and video recording tag information of the corresponding camera according to each result, acquire video training materials from the monitoring system, and retrain the target intelligent algorithm using the video training materials to obtain a new intelligent algorithm.

[0093] The notification unit is used to notify the video management server to update the new intelligent algorithm to all cameras using the target intelligent algorithm.

[0094] Based on the above embodiments, the first data determining unit is specifically used for:

[0095] According to the update instruction, retrieve the camera identifier and time point included in each result correction record from the answer record table used to store the analysis results of the monitoring system;

[0096] Among them, the result correction record refers to the analysis result record that has been corrected in the answer record table, and each result correction record includes at least the camera identifier, time point, intelligent algorithm identifier, analysis result information and correction result information.

[0097] Based on the above embodiments, optionally, the second data determining unit is specifically used for:

[0098] Based on the camera identifier included in each result correction record, the scene type information of the camera corresponding to each correction record is determined by querying the camera scene type table; wherein, each record in the camera scene type table includes at least the camera identifier, the identifier of the intelligent algorithm deployed in the camera, and the scene type information of the camera;

[0099] Based on the time point included in each result correction record, the corresponding video recording tag information for each correction record is determined by querying the video recording category table. The video recording category table is updated based on the data in the answer record table and is used to record the tag information of video recordings in the monitoring system. Each record in the video recording category table includes at least the camera identifier, time range, and video recording tag information.

[0100] Based on the above embodiments, optionally, the material acquisition and training unit is specifically used for:

[0101] For any result, the video recording tag information included in the correction record is used, and video recordings with the same video recording tag information are determined as the first type of video training material according to the video recording category table.

[0102] For any result, the scene type information included in the correction record is used. Based on the camera scene type table, the target camera with the same scene type information is determined, and the video recordings collected by the target camera are used as the second type of video training material.

[0103] The target intelligent algorithm was retrained using the first type of video training material and the second type of video training material respectively, resulting in a new intelligent algorithm.

[0104] Based on the above embodiments, optionally, before notifying the video management server to update the new intelligent algorithm to all cameras using the target intelligent algorithm, the device further includes:

[0105] The verification and judgment module is used to analyze newly generated video recordings using new intelligent algorithms and determine whether the accuracy of the analysis results meets preset conditions.

[0106] The execution module is used to, if so, notify the video management server to update the new intelligent algorithm to all cameras using the target intelligent algorithm.

[0107] Based on the above embodiments, the apparatus may optionally further include:

[0108] The data acquisition module is used to acquire scene type information for any newly added camera in the video surveillance system.

[0109] The algorithm confirmation module is used to determine the intelligent algorithm used by other cameras that have the same camera scene type information as the newly added camera from the camera scene type table.

[0110] The algorithm deployment module is used to select an intelligent algorithm from those used by other cameras and deploy it to the new camera.

[0111] Based on the above embodiments, optionally, the algorithm deployment module is specifically used for:

[0112] If other cameras use the same type of intelligent algorithm, then deploy the intelligent algorithm used by the other cameras to the new camera;

[0113] If other cameras use different types of intelligent algorithms, statistical analysis results of each intelligent algorithm within a preset time period are collected; based on each intelligent algorithm, the video recordings captured by the new camera within the preset time period are analyzed to obtain at least two sampling analysis results; based on the similarity between the reference analysis results and the sampling analysis results corresponding to each intelligent algorithm, an intelligent algorithm is selected and deployed to the new camera.

[0114] The camera intelligent algorithm updating device provided in this disclosure can execute the camera intelligent algorithm updating method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0115] Figure 5 This is a schematic diagram of the structure of an electronic device provided in the fifth embodiment of this disclosure. For example... Figure 5 The structure shown in this embodiment of the electronic device includes: one or more processors 502 and a memory 501; the processors 502 in the electronic device may be one or more. Figure 5 Taking a processor 502 as an example; memory 501 is used to store one or more programs; one or more programs are executed by one or more processors 502, causing one or more processors 502 to implement the camera intelligent algorithm update method as described in any of the embodiments of this disclosure.

[0116] The electronic device may also include an input device 503 and an output device 504.

[0117] The processor 502, memory 501, input device 503, and output device 504 in this electronic device can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0118] The memory 501 in this electronic device serves as a computer-readable storage medium, capable of storing one or more programs. These programs can be software programs, computer-executable programs, or modules, such as the program instructions / modules corresponding to the application control method provided in this embodiment. The processor 502 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 501, thereby implementing the camera intelligent algorithm update method in the above-described method embodiment.

[0119] Memory 501 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, memory 501 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, memory 501 may further include memory remotely located relative to processor 502, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0120] Input device 503 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. Output device 504 may include display devices such as a display screen.

[0121] Furthermore, when one or more programs included in the aforementioned electronic device are executed by one or more processors 502, the programs perform the following operations:

[0122] Obtain scene type information of the camera in the video surveillance system;

[0123] Based on the scene type information of the camera, a scene-type related intelligent algorithm is sent to the camera, enabling the camera to perform business analysis based on the received intelligent algorithm;

[0124] During the use of the video surveillance system, the intelligent algorithm is triggered and updated based on the analysis results data fed back by the camera based on the intelligent algorithm.

[0125] Of course, those skilled in the art will understand that when one or more programs included in the above-described electronic device are executed by one or more processors 502, the programs can also perform related operations in the application control methods provided in any embodiment of this disclosure.

[0126] One embodiment of this disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is used to perform a method for updating a camera intelligent algorithm, the method comprising:

[0127] Obtain scene type information of the camera in the video surveillance system;

[0128] Based on the scene type information of the camera, a scene-type related intelligent algorithm is sent to the camera, enabling the camera to perform business analysis based on the received intelligent algorithm;

[0129] During the use of the video surveillance system, the intelligent algorithm is triggered and updated based on the analysis results data fed back by the camera based on the intelligent algorithm.

[0130] Optionally, when executed by a processor, the program can also be used to perform the methods provided in any embodiment of this disclosure.

[0131] The computer storage medium of this disclosure can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination thereof. The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0132] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0133] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0134] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network (e.g., including local area networks (LANs) or wide area networks (WANs)), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0135] Note that the above description is merely a preferred embodiment and the technical principles employed in this disclosure. Those skilled in the art will understand that this disclosure is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this disclosure. Therefore, although this disclosure has been described in detail through the above embodiments, it is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this disclosure, and the scope of this disclosure is determined by the scope of the appended claims.

Claims

1. A method of updating a camera intelligence algorithm, characterized by, The method comprises: acquiring scene type information of a camera in a video monitoring system; the scene type information at least comprises geographical position information of the camera, service type and resolution of the camera; according to the scene type information of the camera, issuing a scene type related intelligent algorithm to the camera, so that the camera performs service analysis based on the received intelligent algorithm; during use of the video monitoring system, triggering and updating the intelligent algorithm according to analysis result data fed back by the camera based on the intelligent algorithm, comprising: during use of the video monitoring system, if a video management server in the video monitoring system detects that the number of result correction records included in the analysis result data based on the intelligent algorithm reaches a threshold value, issuing an update instruction of the intelligent algorithm to a training server in the video monitoring system; according to the update instruction, acquiring camera identification and time points included in each result correction record; according to the camera identification and time points included in each result correction record, respectively determining scene type information and video recording label information of the camera corresponding to each result correction record; the video recording label information comprises subdivided target type and natural type; according to the scene type information and video recording label information of the camera corresponding to each result correction record, acquiring video training materials from the monitoring system, and retraining the intelligent algorithm using the video training materials to obtain a new intelligent algorithm; informing the video management server to update the new intelligent algorithm to all cameras using the intelligent algorithm.

2. The method of claim 1, wherein, According to the update instruction, acquiring camera identification and time points included in each result correction record, comprising: according to the update instruction, acquiring camera identification and time points included in each result correction record from an answer record table for saving monitoring system analysis result data; wherein the result correction record refers to an analysis result record that has been corrected in the answer record table, and each result correction record at least comprises camera identification, time point, identification of the intelligent algorithm, analysis result information and correction result information.

3. The method of claim 2, wherein, According to the camera identification and time points included in each result correction record, respectively determining scene type information and video recording label information of the camera corresponding to each result correction record, comprising: according to the camera identification included in each result correction record, determining scene type information of the camera corresponding to each correction record by querying a camera scene type table; wherein each record in the camera scene type table at least comprises camera identification, identification of the intelligent algorithm deployed in the camera and scene type information of the camera; according to the time points included in each result correction record, determining video recording label information corresponding to each correction record by querying a recording category table; wherein the recording category table is updated according to data in the answer record table, and is used to record label information of video recording in the monitoring system, and each record in the recording category table at least comprises camera identification, time range and video recording label information.

4. The method of claim 3, wherein, According to the respective corresponding camera scene type information and video recording label information of each result correction record, video training materials are obtained from the monitoring system, and the intelligent algorithm is retrained using the video training materials to obtain a new intelligent algorithm, including: According to the video recording label information corresponding to any result correction record, video recordings with the same video recording label information are determined as first type video training materials according to the recording category table; According to the scene type information corresponding to any result correction record, target cameras with the same scene type information are determined according to the camera scene type table, and video recordings collected by the target cameras are taken as second type video training materials; The intelligent algorithm is retrained using the first type video training materials and the second type video training materials respectively to obtain a new intelligent algorithm.

5. The method of claim 3, wherein, The method further includes: For any new camera of the video monitoring system, scene type information of the new camera is obtained; From the camera scene type table, intelligent algorithms used by other cameras with the same camera scene type information as the new camera are determined; From the intelligent algorithms used by other cameras, an intelligent algorithm is selected and deployed to the new camera.

6. The method of claim 5, wherein, From the intelligent algorithms used by other cameras, an intelligent algorithm is selected and deployed to the new camera, including: If the types of the intelligent algorithms used by the other cameras are the same, the intelligent algorithms used by the other cameras are deployed to the new camera; If the types of the intelligent algorithms used by the other cameras are different, reference analysis results of each type of intelligent algorithm within a preset time period are counted; based on the analysis of the new camera on video recordings collected by the new camera within the preset time period, at least two sampling analysis results are obtained; according to the similarity between the reference analysis result and the sampling analysis result corresponding to each type of intelligent algorithm, an intelligent algorithm is selected and deployed to the new camera.

7. An electronic device, comprising: It includes: One or more processors; Memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the camera intelligent algorithm updating method according to any one of claims 1-6.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the camera intelligent algorithm updating method according to any one of claims 1-6.

Images