Parameter optimization method and device for target recognition and target recognition system

Abstract

The present application discloses a parameter optimization method, device and target recognition system for target recognition, which relates to the field of image processing technology and is used to make the effect of the front-end analysis device and the central end collaboratively performing target event recognition meet user needs. The method includes: obtaining the analysis result of each target event in multiple target events; the analysis result is obtained by the central end performing a true / false alarm analysis on multiple target events, and the analysis result is used to indicate whether the target event is a false alarm; determining the adjustment target of parameter optimization; determining the parameter adjustment strategy corresponding to the adjustment target, and the parameter adjustment strategy includes the adjustment conditions and adjustment direction of the adjustable recognition parameter; the adjustable recognition parameter is the recognition parameter when the front-end analysis device performs target event recognition; analyzing the adjustable recognition parameter in the true / false analysis result, and when the analysis result of the adjustable recognition parameter meets the adjustment condition, controlling the front-end analysis device to update the adjustable recognition parameter according to the adjustment direction.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a parameter optimization method and device for target recognition and a target recognition system. Background Art

[0002] In the intelligent analysis scenario of video images, it is necessary to identify and report target events from images. Due to the different levels of intelligent construction of projects, there are various front-end analysis devices with different detection capabilities on the edge side. When performing intelligent transformation, different front-end analysis devices have great differences in the operation and analysis effects of target recognition algorithms.

[0003] If the detection capability of the front-end analysis device is not enabled, all images are directly connected to the central end for intelligent analysis, and the central end detects the target events in the image and performs target event recognition. If the user has purchased a front-end analysis device, this method does not enable the detection capability of the front-end analysis device at all, which greatly wastes the detection capability of the front-end analysis device. If the detection capability of the front-end analysis device is enabled, the central end analyzes and outputs the final detection results based on the target events reported by the front-end analysis device. Due to the different detection capabilities of different scenarios and different front-end analysis devices, the final detection results output by the central end are not good and cannot meet user requirements.

[0004] Therefore, a method is needed that can both enable the detection capability of the front-end analysis equipment and ensure that the detection effect of the target event can meet the user's requirements. Summary of the invention

[0005] The present application provides a parameter optimization method, device and target recognition system for target recognition, which are used to ensure that the effect of target event recognition performed collaboratively by a front-end analysis device and a central end can meet user needs.

[0006] In order to achieve the above technical objectives, this application adopts the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a parameter optimization method for target recognition, comprising:

[0008] Obtaining the analysis result of each target event among the multiple target events; the multiple target events are reported to the central end by the front-end analysis device; the analysis result is obtained by the central end performing a positive / false positive analysis on the multiple target events, and the analysis result is used to indicate whether the target event is a false positive;

[0009] Determine the adjustment target for parameter optimization;

[0010] Determine the parameter adjustment strategy corresponding to the adjustment target, the parameter adjustment strategy includes the adjustment conditions and adjustment directions of the adjustable identification parameters; the adjustable identification parameters are the identification parameters when the front-end analysis device performs event identification;

[0011] The adjustable identification parameter in the analysis result of each target event in the multiple target events is analyzed, and when the analysis result of the adjustable identification parameter meets the adjustment condition, the front-end analysis device is controlled to update the adjustable identification parameter according to the adjustment direction.

[0012] The technical solution provided by the present application brings at least the following beneficial effects: by obtaining the analysis results of the target events reported by the front-end analysis device by the central end, the adjustable identification parameters are adjusted based on the determined adjustment targets; since the adjustable identification parameters are the identification parameters when the front-end analysis device identifies the target events, and the adjustment targets represent the needs of the users, the effect of the front-end analysis device and the central end cooperating to detect the target events after the adjustment meets the adjustment targets, thus satisfying the user's requirements for the front-end analysis device and the central end to cooperating to detect the target events; at the same time, the front-end analysis device configured by the user is also put into use, without wasting the detection capability of the front-end analysis device; and the entire parameter adjustment process can be performed automatically by the device, without the need for the user to manually adjust the parameters, thus saving the user's operating steps.

[0013] In one possible implementation, the adjustment target includes reducing or increasing the computing power utilization rate of the central end; when the adjustment target includes reducing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area and / or shielding area when the front-end analysis device identifies the target object; the adjustment direction of the detection area includes reducing the detection area, and the adjustment direction of the shielding area includes expanding the shielding area; when the adjustment target includes increasing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area, and the adjustment direction of the detection area includes expanding the detection area.

[0014] In one possible implementation, the adjustment target includes lowering or increasing the detection rate of the target event by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the filtering parameters of the target event by the front-end analysis device; the filtering parameters include at least one of the confidence threshold, object size threshold, detection sensitivity and reporting interval threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the detection rate, the adjustment direction of the filtering parameters includes at least one of reducing the confidence threshold, reducing the object size threshold, reducing the detection sensitivity and shortening the reporting interval threshold; when the adjustment target includes reducing the detection rate, the adjustment direction of the filtering parameters includes at least one of increasing the confidence threshold, expanding the object size threshold, improving the detection sensitivity and increasing the reporting interval threshold.

[0015] In one possible implementation, the adjustment target includes lowering or increasing the positive reporting rate of target events reported by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the confidence threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to increase the confidence threshold; when the adjustment target includes reducing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to reduce the confidence threshold.

[0016] In a possible implementation, when the detection area needs to be reduced, the adjustment condition of the detection area includes any one of the following: the reference area is smaller than the detection area, the first area is smaller than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference, the first ratio is smaller than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein, when the adjustment condition of the detection area includes the case where the reference area is smaller than the detection area, the adjustment direction of the detection area is to reduce the detection area in a direction close to the reference area; or, when the detection area needs to be expanded, the adjustment condition of the detection area includes any one of the following: the first area is larger than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference. Assume that the difference, the first ratio is greater than the second ratio, and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein the analysis result is a positive analysis result or an erroneous analysis result; the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result, the first area is the area occupied by the target object in the original image obtained after the center end performs target event recognition on the original image associated with the positive analysis result, the second area is the area occupied by the target object in the positive analysis result in the detection area of ​​the front-end analysis device, the first ratio is the ratio between the first area and the area of ​​the detection area when the front-end analysis device performs target event recognition, and the second ratio is the ratio between the second area and the area of ​​the detection area when the front-end analysis device performs target event recognition.

[0017] In one possible implementation, when it is necessary to expand the shielding area, the adjustment condition of the shielding area includes the existence of a false alarm area, and the expansion of the shielding area is specifically implemented by merging the false alarm area into the original shielding area to expand the shielding area, and the false alarm area is obtained based on the positive analysis result in the analysis result of the target event and the position of the target object in the erroneous analysis result on the original image associated with the positive analysis result and the erroneous analysis result.

[0018] In a possible implementation, when the object size threshold needs to be reduced, the adjustment condition of the object size threshold includes that the minimum size is less than the object size threshold; or, when the object size threshold needs to be expanded, the adjustment condition of the object size threshold includes that the minimum size is greater than the object size threshold; or, when the detection sensitivity needs to be reduced, the adjustment condition of the detection sensitivity includes that the third ratio is less than a preset ratio; or, when the detection sensitivity needs to be increased, the adjustment condition of the detection sensitivity includes that the third ratio is greater than or equal to the preset ratio; or, when the reporting interval threshold needs to be shortened, the adjustment condition of the reporting interval threshold includes that the reporting interval threshold is greater than the shortest duration. continuation duration; or, in the case where it is necessary to increase the reporting interval threshold, the adjustment condition of the continuation duration threshold includes that the reporting interval threshold is less than the shortest continuation duration; wherein the analysis result is a positive analysis result or an erroneous analysis result; the minimum size is the minimum value of the size of the target object obtained after the center end identifies the original image associated with the positive analysis result; the third ratio is used to characterize the accuracy of the target object that intersects with the boundary of the detection area of ​​the front-end analysis device in the target event reported by the front-end analysis device, and the third ratio is obtained based on the positive analysis result and the erroneous analysis result; the shortest continuation duration is the shortest value of the continuation duration of the target event obtained after the center end identifies the original image associated with the positive analysis result.

[0019] In one possible implementation, when the confidence threshold needs to be reduced, the adjustment condition of the confidence threshold includes that the false alarm rate is greater than or equal to a preset false alarm rate; or, when the confidence threshold needs to be increased, the adjustment condition of the confidence threshold includes that the false alarm rate is less than a preset false alarm rate; wherein the analysis result is a positive analysis result or an erroneous analysis result; and the false alarm rate is the ratio between the number of erroneous analysis results in the analysis results of multiple target events and the number of all analysis results.

[0020] In one possible implementation, the front-end analysis device is controlled to update the adjustable identification parameter according to the adjustment direction, including: when the adjustment target is a single target, based on the degree of influence of each adjustable identification parameter on the adjustment target in the adjustment direction, determining the first weight of the adjustable identification parameter in the adjustment direction; based on the first weight of the adjustable identification parameter, determining the adjustment range of the adjustable identification parameter; controlling the front-end analysis device to update the adjustable identification parameter according to the adjustment direction and the adjustment range.

[0021] In a possible implementation, the front-end analysis device is controlled to update the adjustable identification parameters according to the adjustment direction, including: when the adjustment target is multiple targets, determining the second weight of each single target in the adjustment target, the second weight is used to characterize the importance of the single target to the adjustment target; determining the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and determining the adjustment range of the adjustable identification parameter; based on the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and the second weight of each single target in the adjustment target, determining the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target; based on the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target, determining the adjustment frequency of each adjustable identification parameter involved in the adjustment target; controlling the front-end analysis device to update the adjustable identification parameters according to the adjustment direction, adjustment range, and adjustment frequency.

[0022] In a second aspect, the present application provides a parameter optimization device for target recognition, comprising:

[0023] The transmission module is used to obtain the analysis result of each target event among the multiple target events; the multiple target events are reported to the central end by the front-end analysis device; the analysis result is obtained by the central end performing a positive / false alarm analysis on the multiple target events, and the analysis result is used to indicate whether the target event is a false alarm;

[0024] A processing module, used for determining an adjustment target for parameter optimization;

[0025] The processing module is also used to determine a parameter adjustment strategy corresponding to the adjustment target, the parameter adjustment strategy includes an adjustment condition and an adjustment direction of the adjustable identification parameter; the adjustable identification parameter is an identification parameter when the front-end analysis device performs target event identification;

[0026] The processing module is also used to analyze the adjustable identification parameters in the analysis results of each target event in the multiple target events, and when the analysis results of the adjustable identification parameters meet the adjustment conditions, control the front-end analysis device to update the adjustable identification parameters according to the adjustment direction.

[0027] In one possible implementation, the adjustment target includes reducing or increasing the computing power utilization rate of the central end; when the adjustment target includes reducing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area and / or shielding area when the front-end analysis device identifies the target object; the adjustment direction of the detection area includes reducing the detection area, and the adjustment direction of the shielding area includes expanding the shielding area; when the adjustment target includes increasing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area, and the adjustment direction of the detection area includes expanding the detection area.

[0028] In one possible implementation, the adjustment target includes lowering or increasing the detection rate of the target event by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the filtering parameters of the target event by the front-end analysis device; the filtering parameters include at least one of the confidence threshold, object size threshold, detection sensitivity and reporting interval threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the detection rate, the adjustment direction of the filtering parameters includes at least one of reducing the confidence threshold, reducing the object size threshold, reducing the detection sensitivity and shortening the reporting interval threshold; when the adjustment target includes reducing the detection rate, the adjustment direction of the filtering parameters includes at least one of increasing the confidence threshold, expanding the object size threshold, improving the detection sensitivity and increasing the reporting interval threshold.

[0029] In one possible implementation, the adjustment target includes lowering or increasing the positive reporting rate of target events reported by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the confidence threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to increase the confidence threshold; when the adjustment target includes reducing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to reduce the confidence threshold.

[0030] In a possible implementation, when the detection area needs to be reduced, the adjustment condition of the detection area includes any one of the following: the reference area is smaller than the detection area, the first area is smaller than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference, the first ratio is smaller than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein, when the adjustment condition of the detection area includes the case where the reference area is smaller than the detection area, the adjustment direction of the detection area is to reduce the detection area in a direction close to the reference area; or, when the detection area needs to be expanded, the adjustment condition of the detection area includes any one of the following: the first area is larger than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference. Assume that the difference, the first ratio is greater than the second ratio, and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein the analysis result is a positive analysis result or an erroneous analysis result; the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result, the first area is the area occupied by the target object in the original image obtained after the center end performs target event recognition on the original image associated with the positive analysis result, the second area is the area occupied by the target object in the positive analysis result in the detection area of ​​the front-end analysis device, the first ratio is the ratio between the first area and the area of ​​the detection area when the front-end analysis device performs target event recognition, and the second ratio is the ratio between the second area and the area of ​​the detection area when the front-end analysis device performs target event recognition.

[0031] In one possible implementation, when it is necessary to expand the shielding area, the adjustment condition of the shielding area includes the existence of a false alarm area, and the expansion of the shielding area is specifically implemented by merging the false alarm area into the original shielding area to expand the shielding area, and the false alarm area is obtained based on the positive analysis result in the analysis result of the target event and the position of the target object in the erroneous analysis result on the original image associated with the positive analysis result and the erroneous analysis result.

[0032] In a possible implementation, when the object size threshold needs to be reduced, the adjustment condition of the object size threshold includes that the minimum size is less than the object size threshold; or, when the object size threshold needs to be expanded, the adjustment condition of the object size threshold includes that the minimum size is greater than the object size threshold; or, when the detection sensitivity needs to be reduced, the adjustment condition of the detection sensitivity includes that the third ratio is less than a preset ratio; or, when the detection sensitivity needs to be increased, the adjustment condition of the detection sensitivity includes that the third ratio is greater than or equal to the preset ratio; or, when the reporting interval threshold needs to be shortened, the adjustment condition of the reporting interval threshold includes that the reporting interval threshold is greater than the shortest duration. continuation duration; or, in the case where it is necessary to increase the reporting interval threshold, the adjustment condition of the continuation duration threshold includes that the reporting interval threshold is less than the shortest continuation duration; wherein the analysis result is a positive analysis result or an erroneous analysis result; the minimum size is the minimum value of the size of the target object obtained after the center end identifies the original image associated with the positive analysis result; the third ratio is used to characterize the accuracy of the target object that intersects with the boundary of the detection area of ​​the front-end analysis device in the target event reported by the front-end analysis device, and the third ratio is obtained based on the positive analysis result and the erroneous analysis result; the shortest continuation duration is the shortest value of the continuation duration of the target event obtained after the center end identifies the original image associated with the positive analysis result.

[0033] In one possible implementation, when the confidence threshold needs to be reduced, the adjustment condition of the confidence threshold includes that the false alarm rate is greater than or equal to a preset false alarm rate; or, when the confidence threshold needs to be increased, the adjustment condition of the confidence threshold includes that the false alarm rate is less than a preset false alarm rate; wherein the analysis result is a positive analysis result or an erroneous analysis result; and the false alarm rate is the ratio between the number of erroneous analysis results in the analysis results of multiple target events and the number of all analysis results.

[0034] In one possible implementation, the processing module is specifically used to: when the adjustment target is a single target, determine the first weight of the adjustable identification parameter in the adjustment direction based on the degree of influence of each adjustable identification parameter on the adjustment target in the adjustment direction; determine the adjustment range of the adjustable identification parameter based on the first weight of the adjustable identification parameter; and control the front-end analysis device to update the adjustable identification parameter according to the adjustment direction and adjustment range.

[0035] In one possible implementation, the processing module is specifically used to: when the adjustment target is multiple targets, determine the second weight of each single target in the adjustment target, the second weight is used to characterize the importance of the single target to the adjustment target; determine the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and determine the adjustment range of the adjustable identification parameter; based on the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and the second weight of each single target in the adjustment target, determine the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target; based on the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target, determine the adjustment frequency of each adjustable identification parameter involved in the adjustment target; control the front-end analysis device to update the adjustable identification parameter according to the adjustment direction, adjustment range, and adjustment frequency.

[0036] In a third aspect, the present application provides a target recognition system, including: an electronic device, a front-end analysis device, and a central end;

[0037] The front-end analysis device is used to detect target events in the image of the video file and report the detected multiple target events to the central end;

[0038] The center end is used to receive multiple target events reported by the front-end analysis device, and perform positive / false positive analysis on the multiple target events to obtain the analysis result of each target event in the multiple target events; the analysis result is used to indicate whether the target event is a false positive;

[0039] The electronic device is used to execute any one of the parameter optimization methods for target recognition provided in the first aspect above.

[0040] In a fourth aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any one of the parameter optimization methods for target identification provided in the first aspect.

[0041] In a fifth aspect, the present application provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement any one of the parameter optimization methods for target identification provided in the first aspect above.

[0042] In a sixth aspect, the present application provides a computer program product, comprising computer instructions, which, when executed by a processor, implement any one of the parameter optimization methods for target identification provided in the first aspect above.

[0043] For the specific description of the second to sixth aspects and their various implementations in the present application, reference may be made to the detailed description in the first aspect and its various implementations; and for the beneficial effects of the second to sixth aspects and their various implementations, reference may be made to the beneficial effects analysis in the first aspect and its various implementations, which will not be repeated here.

[0044] These and other aspects of the present application will become more apparent from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the structure of a target recognition system applicable to a parameter optimization method for target recognition provided in an embodiment of the present application;

[0046] Figure 2 A schematic diagram of the structure of a computing device provided in an embodiment of the present application;

[0047] Figure 3 A flow chart of a parameter optimization method for target identification provided in an embodiment of the present application;

[0048] Figure 4 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 1 ;

[0049] Figure 5 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 2 ;

[0050] Figure 6 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 3 ;

[0051] Figure 7 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 4 ;

[0052] Figure 8 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 5 ;

[0053] Fig. 9 Schematic diagram of an application scenario of a parameter optimization method for target recognition provided in an embodiment of the present application Figure 6 ;

[0054] Fig.10 A logical schematic diagram of a parameter optimization method for target identification provided in an embodiment of the present application;

[0055] Fig.11A schematic diagram of the structure of a parameter optimization device for target identification provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way. The terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, "multiple" means two or more.

[0058] As mentioned in the background technology, in the intelligent analysis scenario of video images, it is necessary to identify and report target events from the image. If the detection capability of the front-end analysis device is not enabled, the configured front-end analysis device will be wasted, and the central end will need to process a large amount of data. When the front-end analysis device and the central end work together to identify target events, the detection capabilities of different scenarios and different front-end analysis devices are different, resulting in the final detection result output by the central end being poor and failing to meet user requirements.

[0059] Therefore, a method is needed that can both enable the detection capability of the front-end analysis equipment and ensure that the detection effect of the target event can meet the user's requirements.

[0060] In this regard, the embodiment of the present application provides a parameter optimization method for target identification. After the central end obtains the target event reported by the front-end analysis device, it performs a true / false alarm analysis to obtain the analysis result of each target event. The present application determines the adjustment target of parameter optimization and the parameter adjustment strategy corresponding to the adjustment target, that is, it determines the effect to be achieved by adjusting the front-end analysis device and the central end to collaboratively identify the target event, and how to adjust the parameters of the front-end analysis device when identifying the target event. Finally, based on the adjustable identification parameters in the analysis results of these target events, it is determined whether to adjust the adjustable identification parameters and control the front-end analysis device to adjust the adjustable identification parameters. The entire parameter adjustment process is performed by the device itself to adjust the parameters when the front-end analysis device identifies the target event, and the user does not need to manually adjust the parameters, which can save the user's operation steps. In addition, the adjustable identification parameters are the identification parameters when the front-end analysis device identifies the target event. The adjustment target represents the user's needs. The effect of the front-end analysis device and the central end collaboratively detecting the target event after adjustment meets the adjustment target, which also meets the user's requirements for the front-end analysis device and the central end to collaboratively detect the target event. At the same time, the front-end analysis device configured by the user is also put into use, and the detection capability of the front-end analysis device is not wasted.

[0061] Please refer to Figure 1 , which shows a schematic diagram of the structure of a target recognition system to which the parameter optimization method for target recognition provided by the present application is applicable. Figure 1 As shown, the target recognition system 1 may include: a front-end analysis device 10 , a central terminal 20 and an electronic device 30 .

[0062] Among them, the electronic device 30 establishes communication connection with the front-end analysis device 10 and the central end 20 respectively, and the front-end analysis device 10 establishes communication connection with the central end 20. It should be understood that the connection method can be a wireless connection, such as a Bluetooth connection, a wireless fidelity (Wi-Fi) connection, etc.; or, the connection method can also be a wired connection, such as an optical fiber connection, etc., which is not limited. Exemplarily, the front-end analysis device 10 and the central end 20 can be connected to the Internet through a router, thereby realizing the communication connection between the front-end analysis device 10 and the central end 20.

[0063] In some embodiments, the front-end analysis device 10 is used to receive images collected by an image acquisition device and perform target event recognition on the images. The image acquisition device can be any device that can collect images and videos and transmit video files and images to the front-end analysis device 10, such as a camera, a still camera, a ball camera, etc.

[0064] In some embodiments, the front-end analysis device 10 can be integrated with an image acquisition device, or the front-end analysis device 10 is connected to the image acquisition device, the image acquisition device is used to acquire images, and the front-end analysis device 10 is used to detect whether a target event occurs in the image. For example, a plurality of image acquisition devices are set up in the management area, and these image acquisition devices can capture scenes in the management area. The image acquisition devices send the acquired video files to the front-end analysis device 10, and then the front-end analysis device 10 detects the images in the video files to determine whether a target event occurs. For another example, the front-end analysis device 10 can be a camera with target event detection capabilities.

[0065] In some embodiments, the front-end analysis device 10 is further configured to report the detected target event to the central end 20 .

[0066] In some embodiments, the front-end analysis device 10 may include a processor. The processor is used to identify target events in the image of the video file. The processor may be a central processing unit (CPU), a general-purpose processor network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller, a programmable logic device (PLD), or any combination thereof. The processor may also be other devices with processing functions, such as circuits, devices, or software modules, and the embodiments of the present application do not impose any restrictions on this.

[0067] Specifically, the front-end analysis device 10 may run a target event recognition algorithm to recognize the target event in the image.

[0068] Optionally, the front-end analysis device 10 may further include a memory for storing images in a video file from the image acquisition device, and then the processor may search the image in the video file from the memory to perform target event recognition on the image in the video file.

[0069] In some embodiments, the central end 20 is used to receive target events reported by the front-end analysis device 10, perform true / false alarm analysis on each target event, and obtain an analysis result for each target event, which is used to indicate whether the target event is a false alarm.

[0070] In some embodiments, the electronic device 30 is used to determine the adjustment target of parameter optimization, determine the parameter adjustment strategy corresponding to the adjustment target, and the parameter adjustment strategy includes the adjustment conditions and adjustment direction of the adjustable identification parameters. The adjustable identification parameters are the identification parameters when the front-end analysis device identifies the target event. The electronic device 30 is also used to analyze the adjustable identification parameters in the analysis results of each target event, and when the analysis results of the adjustable identification parameters meet the adjustment conditions, control the front-end analysis device to update the adjustable identification parameters according to the adjustment direction to ensure that the front-end analysis device 10 and the central end 20 can achieve the optimal target event detection effect when they collaborate to identify the target event.

[0071] In some embodiments, the user can pre-configure the video file, the front-end analysis device 10 identifies each image in the video file, and the central end 20 will also synchronously identify the target event for each image in the video file. Then, the electronic device 30 can compare the target event detected by the front-end analysis device 10 with the target event detected by the central end 20, determine the analysis result of each target event reported by the front-end analysis device 10, and finally adjust the recognition parameters of the front-end analysis device 10 when identifying the target event based on the analysis results of these target events.

[0072] In some embodiments, the user can pre-set an adjustment period, and the front-end analysis device 10 performs target event identification on each image received within the adjustment period, and the central end 20 also synchronously performs target event identification on each image. Then, the electronic device 30 can compare the target events detected by the front-end analysis device 10 with the target events detected by the central end 20, determine the analysis results of each target event reported by the front-end analysis device 10, and finally adjust the identification parameters of the front-end analysis device 10 when performing target event identification based on the analysis results of these target events.

[0073] In some embodiments, the target recognition system 1 may also include a prompt device. After the optimization of the recognition parameters of the front-end analysis device 10 is completed, the central end 20 and the front-end analysis device 20 cooperate to identify the target event, and send the final detection result of the target event to the prompt device, and the prompt device displays the final detection result of the target event to the user. Exemplarily, the prompt device can be a display, a speaker, etc., and the embodiment of the present application does not specifically limit the implementation method of the prompt device. For example, the front-end analysis device 10 reports the target event A to the central end 20, and the central end 20 identifies the original image associated with the target event A and determines that the target event A is a false alarm event. Then the central end 20 will not control the prompt device to display the target event A at this time. For another example, the front-end analysis device 10 reports the target event B to the central end 20, and the central end 20 identifies the original image associated with the target event B and determines that the target event B is a positive alarm event. Then the central end 20 controls the prompt device to display the target event B.

[0074] In some embodiments, the target recognition system 1 may include one or more front-end analysis devices 10 .

[0075] In some embodiments, the front-end analysis device 10 may be a separate computing device, a server cluster, or an image acquisition device with target event detection capability.

[0076] In some embodiments, the central end 20 may be a single server or a server cluster, or the central end 20 may be a terminal device with target event detection capability, such as a personal computer (PC), a notebook computer, a mobile device, a tablet computer, a laptop computer, etc. The embodiments of the present application do not limit the specific form of the central end 20.

[0077] In some embodiments, the electronic device 30 may be a single server or a server cluster, or the electronic device 30 may be a device having a parameter analysis function and a function of adjusting parameters of the front-end analysis device 10 when identifying a target event, such as a personal computer (PC), a notebook computer, a mobile device, a tablet computer, a laptop computer, etc. The embodiments of the present application do not limit the specific form of the electronic device 30.

[0078] In some embodiments, the central terminal 20 and the electronic device 30 can be as follows Figure 1 As shown in FIG. 2 , two independent devices are shown; alternatively, the central terminal 20 and the electronic device 30 may be integrated together.

[0079] The hardware structure of the front-end analysis device 10, the central terminal 20 and the electronic device 30 includes: Figure 2The computing device shown in FIG. Figure 2 Taking the computing device shown as an example, the hardware structure of the front-end analysis device 10, the central terminal 20 and the electronic device 30 is introduced.

[0080] like Figure 2 As shown, the computing device may include a processor 401, a memory 402, a communication interface 403, and a bus 404. The processor 401, the memory 402, and the communication interface 403 may be connected via a bus 404.

[0081] The processor 401 is the control center of the computing device, and can be a processor or a general term for multiple processing elements. For example, the processor 401 can be a general-purpose central processing unit (CPU) or other general-purpose processors. Among them, the general-purpose processor can be a microprocessor or any conventional processor.

[0082] As an embodiment, the processor 401 may include one or more CPUs, such as Figure 2 CPU 0 and CPU 1 are shown in .

[0083] The memory 402 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0084] In a possible implementation, the memory 402 may exist independently of the processor 401, and the memory 402 may be connected to the processor 401 via a bus 404 to store instructions or program codes. When the processor 401 calls and executes the instructions or program codes stored in the memory 402, the model deployment method provided in the embodiment of the present application can be implemented.

[0085] In another possible implementation, the memory 402 may also be integrated with the processor 401 .

[0086] The communication interface 403 is used for the computing device to connect with other devices through a communication network, which may be Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc. The communication interface 403 may include a receiving unit for receiving data and a sending unit for sending data.

[0087] The bus 404 may be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus, or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0088] It should be pointed out that Figure 2 The structure shown in the figure does not constitute a limitation on the computing device, except Figure 2 In addition to the components shown, the computing device may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0089] The implementation of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0090] The parameter optimization method for target recognition provided in the embodiment of the present application can be executed by the electronic device 30 in the target recognition system 1.

[0091] like Figure 3 As shown, the embodiment of the present application provides a parameter optimization method for target recognition, which includes the following steps:

[0092] S101. Obtain analysis results of each target event among multiple target events.

[0093] Among them, multiple target events are reported to the central end by the front-end analysis device, and the analysis results of the target events are obtained by the central end by performing a true / false alarm analysis on the multiple target events. The analysis results of the target events are used to indicate whether the target events are false alarms.

[0094] In some embodiments, the center end performs a positive / false alarm analysis on multiple target events to obtain the analysis results of the target events, which can be specifically implemented as follows: the center end also performs target event recognition on the original image associated with the target event reported by the front-end analysis device to obtain the target event recognition result corresponding to the original image; if the recognition result recognized by the center end matches the target event reported by the front-end analysis device, it is determined that the target event reported by the front-end analysis device is a positive alarm event, and the center end performs a positive / false alarm analysis on the target event to obtain the analysis result of the target event as a positive analysis result; if the recognition result recognized by the center end does not match the target event reported by the front-end analysis device, it is determined that the target event reported by the front-end analysis device is a false alarm event, and the center end performs a positive / false alarm analysis on the target event to obtain the analysis result of the target event as a false analysis result. Among them, "matching" can be understood as that the area occupied by the target object in the target event reported by the front-end analysis device on the original image and the area occupied by the target object identified by the center end from the original image reach a preset threshold, then it can be considered that the target event detected by the front-end analysis device matches the recognition result recognized by the center end, otherwise it does not match.

[0095] For example, the target event is garbage overflowing from a trash can. If within the adjustment period of 02:10:30-03:10:30, at 02:15:30, the front-end analysis device reports target event 1, at 02:18:03, the front-end analysis device reports target event 2, at 02:25:16, the front-end analysis device reports target event 3, at 02:34:57, the front-end analysis device reports target event 4, at 02:52:31, the front-end analysis device reports target event 5, at 02:58:17, the front-end analysis device reports target event 6, and at 02:25:16, the front-end analysis device reports target event 7. At 3:04:42, the front-end analysis device reported target event 7. The central end performed target event recognition on the original images associated with target events 1 to 7, and obtained recognition results 1 to 7. If recognition results 1, 3, and 6 all indicate that there is no garbage overflow from the trash can in the image, and recognition results 2, 4, 5, and 7 all indicate that there is garbage overflow from the trash can in the image, then the central end can determine that target events 1, 3, and 6 are all false alarm events, and target events 2, 4, 5, and 7 are all positive alarm events.

[0096] In some embodiments, the user can pre-set an adjustment cycle. The central end receives multiple target events reported by the front-end analysis device within each adjustment cycle, and simultaneously identifies the target event for each frame of video image identified by the front-end analysis device. The identification results of the front-end analysis device and the central end are combined to facilitate the electronic device to adjust the identification parameters when the front-end analysis device identifies the target event.

[0097] For example, the user sets one hour as an adjustment cycle. Assuming that the start time of detecting the target event is 02:10:30, then between 02:10:30-03:10:30, the center receives multiple target events reported by the front-end analysis device, and performs true / false alarm analysis on the target events. Finally, the recognition parameters of the front-end analysis device when identifying the target event are adjusted according to the analysis results of each target event.

[0098] In other embodiments, a user can input a video into a front-end analysis device and a central end at the same time. The front-end analysis device and the central end synchronously identify target events for each frame of the video. Combining the recognition results of the front-end analysis device and the central end, the electronic device can facilitate the subsequent adjustment of recognition parameters when the front-end analysis device identifies the target event.

[0099] In some embodiments, when the central end and the front-end analysis device simultaneously identify target events for a video, the central end can detect target events that have not been detected by the front-end analysis device. This is a missed event of the front-end analysis device. In the subsequent process, the parameters of the front-end analysis device can be adjusted based on the analysis results of the missed events, true positive events, and false positive events in the video.

[0100] S102: Determine an adjustment target for parameter optimization, and determine a parameter adjustment strategy corresponding to the adjustment target.

[0101] The parameter adjustment strategy includes the adjustment conditions and adjustment directions of the adjustable identification parameters; the adjustable identification parameters are the identification parameters used by the front-end analysis device when performing target event identification.

[0102] In some embodiments, the adjustment target may include one or more of the following: reducing the utilization rate of the computing power of the central end, increasing the utilization rate of the computing power of the central end, increasing the detection rate of the target event by the front-end analysis device, reducing the detection rate of the target event by the front-end analysis device, increasing the positive reporting rate of the target event reported by the front-end analysis device, and reducing the positive reporting rate of the target event reported by the front-end analysis device. The adjustment target can be divided into a single target and multiple targets. A single target means that the adjustment target only includes one of the above targets, and a multiple target means that the adjustment target includes at least two of the above targets.

[0103] When the adjustment target includes reducing the utilization rate of the computing power at the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area and / or shielding area when the front-end analysis device identifies the target object; the adjustment direction of the detection area includes reducing the detection area, and the adjustment direction of the shielding area includes expanding the shielding area.

[0104] The detection area is used to indicate the area in the image where the front-end analysis device identifies target events. Reducing the detection area means that the detection range of the front-end analysis device becomes smaller, and the number of target events detected by the front-end analysis device in the detection area will also be relatively reduced, and the number of target events to be analyzed by the center will also be reduced. Therefore, reducing the detection area will reduce the computing power utilization rate of the center.

[0105] The shielded area is used to indicate the area in the image where the front-end analysis device does not identify target events. Expanding the shielded area means that the front-end analysis device does not need to detect a larger range, and the number of target events detected by the front-end analysis device will also be relatively reduced, and the number of target events to be analyzed by the center will also be reduced. Therefore, expanding the shielded area will reduce the utilization rate of the center's computing power.

[0106] When the adjustment goal includes improving the utilization rate of the computing power at the center end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area, and the adjustment direction of the detection area includes expanding the detection area.

[0107] Expanding the detection area means that the detection range of the front-end analysis device becomes larger, and the number of target events detected by the front-end analysis device in the detection area will also increase relatively, and the number of target events to be analyzed by the center will also increase. Therefore, expanding the detection area will increase the utilization rate of the center's computing power.

[0108] In the case where the adjustment target includes lowering or increasing the detection rate of the target event by the front-end analysis device, the adjustable identification parameters in the parameter adjustment strategy include the filtering parameters of the target event by the front-end analysis device; the filtering parameters include at least one of the confidence threshold, object size threshold, detection sensitivity, and reporting interval threshold when the front-end analysis device identifies the target object. In the case where the adjustment target includes increasing the detection rate of the target event by the front-end analysis device, the adjustment direction of the filtering parameters includes at least one of reducing the confidence threshold, reducing the object size threshold, reducing the detection sensitivity, and shortening the reporting interval threshold. In the case where the adjustment target includes lowering the detection rate of the target event by the front-end analysis device, the adjustment direction of the filtering parameters includes at least one of increasing the confidence threshold, expanding the object size threshold, improving the detection sensitivity, and increasing the reporting interval threshold.

[0109] The confidence threshold is used to indicate the confidence threshold when the front-end analysis device filters the target event. The confidence of the target object identified in the target event reported by the front-end analysis device is greater than or equal to the confidence threshold. The confidence is used to characterize the degree of credibility of the target object identified in the target event reported by the front-end analysis device as considered by the front-end analysis device. When the confidence threshold is reduced, the front-end analysis device will also report the target events that would not have been reported before, and the number of target events detected by the front-end analysis device will increase relatively, which means that the detection rate of the target event by the front-end analysis device is improved. Correspondingly, when the confidence threshold is increased, the detection rate of the target event by the front-end analysis device will be reduced.

[0110] The object size threshold is used to indicate the size threshold of the target objects in the target events filtered by the front-end analysis device. The size of the target objects in the target events reported by the front-end analysis device is greater than or equal to the object size threshold. When the object size threshold is reduced, the front-end analysis device will identify smaller objects that would not have been detected before as target objects. The number of target objects increases, and the number of target events also increases relatively, which means that the detection rate of the target events by the front-end analysis device is improved. Correspondingly, when the object size threshold is expanded, the detection rate of the target events by the front-end analysis device will be reduced.

[0111] Detection sensitivity is used to indicate the sensitivity threshold that triggers the front-end analysis device to detect the target event. By reducing the detection sensitivity, more objects can trigger the front-end analysis device to detect the target event, which means that the detection rate of the front-end analysis device for the target event is improved. Correspondingly, when the object size threshold is increased, the detection rate of the front-end analysis device for the target event will be reduced.

[0112] The reporting interval threshold is used to indicate the duration of the target event detected by the front-end analysis device. The duration of the target event reported by the front-end analysis device is greater than or equal to the reporting interval threshold. When the reporting interval threshold is shortened, some target events with short duration will be detected, and the number of target events detected by the front-end analysis device will increase, which means that the detection rate of the target event by the front-end analysis device is improved. Correspondingly, when the reporting interval threshold is increased, the detection rate of the target event by the front-end analysis device will be reduced.

[0113] In the case where the adjustment target includes lowering or increasing the positive reporting rate of target events reported by the front-end analysis device, the adjustable identification parameters in the parameter adjustment strategy include the confidence threshold when the front-end analysis device identifies the target object. In the case where the adjustment target includes increasing the positive reporting rate of target events reported by the front-end analysis device, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to increase the confidence threshold; in the case where the adjustment target includes lowering the positive reporting rate of target events reported by the front-end analysis device, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to reduce the confidence threshold.

[0114] By increasing the confidence threshold, only target events that the front-end analysis device considers to be highly credible will be reported, and the probability that the target events reported by the front-end analysis device are positive events will increase, that is, the positive reporting rate of the target events reported by the front-end analysis device will be increased. Correspondingly, when the confidence threshold is reduced, the positive reporting rate of the target events reported by the front-end analysis device will be reduced.

[0115] Based on the above description, the adjustment conditions corresponding to the adjustment directions of different adjustable recognition parameters are as follows:

[0116] In the case where it is necessary to reduce the detection area, the adjustment conditions of the detection area include any one of the following: the reference area is smaller than the detection area, the first area is smaller than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference, the first ratio is smaller than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein, in the case where the adjustment conditions of the detection area include the case where the reference area is smaller than the detection area, the adjustment direction of the detection area is to reduce the detection area in the direction closer to the reference area; or, in the case where it is necessary to expand the detection area, the adjustment conditions of the detection area include any one of the following: the first area is larger than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference, the first The ratio is greater than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein the analysis result is a positive analysis result or an erroneous analysis result; the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result, the first area is the area occupied by the target object in the original image obtained after the center end performs target event recognition on the original image associated with the positive analysis result, the second area is the area occupied by the target object in the positive analysis result in the detection area of ​​the front-end analysis device, the first ratio is the ratio between the first area and the area of ​​the detection area when the front-end analysis device performs target event recognition, and the second ratio is the ratio between the second area and the area of ​​the detection area when the front-end analysis device performs target event recognition.

[0117] In some embodiments, the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result. Specifically, it can be: the minimum rectangular area occupied by the target object in the positive analysis result on the original image associated with the positive analysis result is determined as the reference area, or the minimum rectangular area is deformed to obtain the reference area.

[0118] For example, the pixel coordinates of the target object located in the upper left corner of all the target objects in the analysis results are (x1, y1), and the pixel coordinates of the target object located in the lower right corner are (x2, y2). Then the reference area is the area surrounded by the four points (x1, y1), (x2, y1), (x1, y2) and (x2, y2).

[0119] For example, Figure 4 , Figure 5 and Figure 6 The figure shows the original images associated with the three target events reported by the front-end analysis device. The solid line frame in each image represents the person. The analysis results of the three target events are all positive analysis results. The reference area determined by the electronic device can be as follows: Figure 7 As shown, the reference area is Figure 4 , Figure 5 and Figure 6 The minimum bounding rectangle of the three solid-line boxes in the same image, that is, the reference area is Figure 7 The area indicated by the dashed box.

[0120] In some embodiments, shrinking the detection area in the direction close to the reference area can be specifically implemented by: shrinking each side of the detection area by a preset length so that the reduced detection area is closer to the reference area, or shortening the side of the detection area close to the target object in the more erroneous analysis results to shrink the detection area. In actual applications, other methods can also be used to shrink the detection area, and this application does not limit the specific method of shrinking the detection area.

[0121] The reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result, which means that the result detected in the reference area is more reliable. The closer the detection area is to the reference area, the better the detection effect of the target event is.

[0122] If the first area is larger than the second area, it means that too few target objects are detected in the detection area corresponding to the front-end analysis device, and there are actually target objects outside the detection area that have not been detected. The size of the detection area may be small, and the detection result of the front-end analysis device may not be accurate enough relative to the detection result of the center end. It is necessary to expand the size of the detection area to detect more target objects and get close to the detection effect of the center end. If the first area is smaller than the second area, it means that the size of the detection area corresponding to the front-end analysis device may be large, so too many target objects are detected, and it is necessary to reduce the size of the detection area so that the detection result of the front-end analysis device is close to the detection result of the center end.

[0123] When the detection area is specifically expanded or reduced, the greater the absolute value of the difference between the first area and the second area, the greater the expansion or reduction. For example, if the absolute value of the difference between the first area and the second area is 10%, the expansion is 10% of the original length; if the absolute value of the difference between the first area and the second area is 5%, the expansion is 5% of the original length. For another example, the absolute value of the difference between the first area and the second area in different ranges corresponds to different expansion or reduction ranges. When the absolute value of the difference between the first area and the second area falls between [0, 10%), the reduction is 5% of the original length; when the absolute value of the difference between the first area and the second area falls between [10%, 20%), the reduction is 10% of the original length; when the absolute value of the difference between the first area and the second area falls between [20%, 30%), the reduction is 15% of the original length.

[0124] For another example, the detection area is expanded by increasing the side length of the detection area by a preset length. Assuming that the original detection area is 500 pixels long, 300 pixels wide, and the preset length is 20 pixels, the resolution of the expanded detection area is 520×320.

[0125] For example, Figure 8 As shown in (a) in the figure, the detection area is the area shown in the solid line frame. Part of the target object is within the detection area, and the other part is outside the detection area. The detection frame corresponding to the target object reported by the front-end analysis device is shown in the dotted line frame. The detection frame corresponding to the target object detected by the center end can be shown as Figure 8 As shown in (b), it can be seen that the detection area of ​​the front-end analysis device should be expanded to the right. Therefore, when adjusting the detection area of ​​the front-end analysis device, the length of the detection area can be extended to the right by a preset length to expand the detection area of ​​the front-end analysis device.

[0126] Users can expand the specific method of expanding or reducing the detection area based on application requirements. This application does not limit the specific method of expanding or reducing the detection area.

[0127] When the adjustment condition involves the first ratio and the second ratio, the process of determining whether the adjustment condition is satisfied is similar to that of the first area and the second area, and will not be described in detail here.

[0128] In this way, by determining the first area and the second area, or determining the first ratio and the second ratio, it can be determined whether the adjustable identification parameters of the front-end analysis device meet the adjustment requirements, and then the parameters of the target event identification of the front-end analysis device are adjusted when the adjustment requirements are met, so that the effect of the front-end analysis device in detecting the target event is closer to the detection effect of the central end.

[0129] When it is necessary to expand the shielding area, the adjustment conditions of the shielding area include the existence of a false alarm area. The expansion of the shielding area is specifically achieved by merging the false alarm area into the original shielding area to expand the shielding area. The false alarm area is obtained based on the position of the target object in the positive analysis result and the false analysis result in the analysis result of the target event on the original image associated with the positive analysis result and the false analysis result.

[0130] The false alarm area is obtained based on the position of the target object in the positive analysis result and the false analysis result on the original image associated with the positive analysis result and the false analysis result, and can be specifically implemented as follows: based on the outer rectangular area formed by the target object in the false analysis result, the false alarm area is determined, or the image is divided into multiple small areas, based on the position of the target object in the positive analysis result and the false analysis result on the original image associated with the positive analysis result and the false analysis result, the false alarm rate of the target event in each small area is determined, and the area with a false alarm rate greater than or equal to the false alarm rate threshold is determined as the false alarm area, or the area with a positive alarm rate less than or equal to the positive alarm rate threshold is determined as the false alarm area. Users can expand other ways of determining false alarm areas based on actual applications.

[0131] For example, the pixel coordinates of the target object at the upper left corner of all false alarm events are (x1, y1), and the pixel coordinates of the target object at the lower right corner are (x2, y2). Then the false alarm area is the area surrounded by the four points (x1, y1), (x2, y1), (x1, y2) and (x2, y2). Fig. 9 As shown, the area enclosed by the dotted box is the circumscribed rectangular area formed by the positions of the target objects in all false alarm events within a period of time. There has never been a positive alarm event in this area, which means that the target events detected in this area are likely to be wrong, and it may not be necessary to perform target event detection in this area. Therefore, this area can be determined as a false alarm area, and this false alarm area can be incorporated into the original shielding area to expand the scope of the shielding area. The area where the front-end analysis device does not perform target event detection is larger, which reduces the computing power utilization rate of the center end. In addition, the shielding area is expanded based on the false alarm area, which can make the detection result of the target event by the front-end analysis device more reliable.

[0132] When the shielding area is actually expanded, the shielding area may be expanded by merging the false alarm area, increasing the side length of the shielding area, etc. When increasing the side length of the shielding area, the side length of the shielding area may be increased according to a preset ratio, or may be increased according to a preset increment....

[0133] When the confidence threshold needs to be reduced, the adjustment condition of the confidence threshold includes that the false alarm rate is greater than or equal to the preset false alarm rate; or, when the confidence threshold needs to be increased, the adjustment condition of the confidence threshold includes that the false alarm rate is less than the preset false alarm rate; wherein, the analysis result is a positive analysis result or an erroneous analysis result; the false alarm rate is the ratio between the number of erroneous analysis results in the analysis results of multiple target events and the number of all analysis results.

[0134] The adjustment range of the confidence threshold can be positively correlated with the difference between the false alarm rate of the front-end analysis device and the preset false alarm rate. The larger the difference, the larger the expansion range, and the smaller the difference, the smaller the expansion range. When the confidence threshold needs to be increased, the confidence threshold can be increased according to a preset ratio, or according to a preset variable, or according to a preset adjustment variable corresponding to the difference between the false alarm rate and the preset false alarm rate, ..., the user can design a suitable increase method based on the actual application situation, and the method of reducing the confidence threshold is the same, which will not be repeated here.

[0135] As an example, the center receives 15 target events reported by the front-end analysis device during the adjustment period, of which 12 are positive alarm events and 3 are false alarm events, and the false alarm rate is 20%. Assuming the preset false alarm rate is 15%, the false alarm rate is greater than the preset false alarm rate. If the adjustment direction in the parameter adjustment strategy corresponding to the adjustment target is to reduce the confidence threshold, then the confidence threshold meets the adjustment conditions at this time and needs to be reduced. The difference between the false alarm rate and the preset false alarm rate is 5%. Assuming the corresponding adjustment range is 5%, the original confidence threshold is a, then the reduced confidence threshold can be 0.95a.

[0136] As another example, the central end receives 15 target events reported by the front-end analysis device during the adjustment period, of which 12 are positive alarm events and 1 is a false alarm event, and the false alarm rate is 6.6%. Assuming that the preset false alarm rate is 15%, the false alarm rate is less than the preset false alarm rate. If the adjustment direction in the parameter adjustment strategy corresponding to the adjustment target is to increase the confidence threshold, then the confidence threshold meets the adjustment conditions at this time and needs to be increased. The difference between the false alarm rate and the preset false alarm rate is 8.4%. Assuming that the corresponding adjustment range is 8%, and the original confidence threshold is a, the increased confidence threshold is 1.08a.

[0137] When the false alarm rate is greater than or equal to the preset false alarm rate, it means that the confidence threshold is too low, resulting in false alarm events being triggered and reported. By increasing the confidence threshold, the detection rate of the target event by the front-end analysis device can be reduced, and the positive alarm rate of the target event by the front-end analysis device can be increased. When the false alarm rate is less than the preset false alarm rate, it means that the confidence threshold is too high, and missed alarms may occur. By lowering the confidence threshold, the detection rate of the target event by the front-end analysis device can be increased, and the positive alarm rate of the target event by the front-end analysis device can be reduced.

[0138] In some embodiments, when the electronic device increases the confidence threshold, it can determine the expansion range of the confidence threshold according to the difference between the first detection confidence and the second detection confidence. The first detection confidence is the confidence of the target object detected by the front-end analysis device in the false alarm event, and the second detection confidence is the confidence of the target object obtained by the center end after the target object is identified in the original image associated with the false alarm event.

[0139] Specifically, the electronic device determines the average value of the difference between the first detection confidence and the second detection confidence corresponding to all false alarm events, and finally increases the confidence threshold of the front-end analysis device according to the average value. The average value and the increase in the confidence threshold can be positively correlated. The larger the average value, the greater the increase in the confidence threshold. The electronic device can increase the confidence threshold according to a preset ratio, or expand the confidence threshold according to a fixed increase variable, or increase the confidence threshold according to the increase variable corresponding to the average value. This application does not limit the specific method of increasing the confidence threshold.

[0140] For example, assuming that in the false alarm events reported by the front-end analysis device, in the original image associated with the false alarm events, the first detection confidence of the target object detected by the front-end analysis device is 66%, and the second detection confidence of the target object detected by the central end is 88%. During the adjustment period, the front-end analysis device reported 5 false alarm events, and each false alarm event corresponds to a first detection confidence and a second detection confidence. Assuming that the average value of the difference between the first detection confidence and the second detection confidence corresponding to all false alarm events is 10%, the corresponding adjustment range is determined to be 10%, and the confidence threshold before the increase is b, then the adjusted confidence threshold is 1.1b.

[0141] The larger the average value is, the larger the difference between the first detection confidence and the second detection confidence corresponding to the false alarm event is, that is, the credibility of the detection result of the front-end analysis device deviates greatly from the credibility of the detection result of the central end, and the accuracy of the target event detected by the front-end analysis device is not high. In this case, the confidence threshold of the front-end analysis device should be increased. After the adjustment, the front-end analysis device will report the target event only when it believes that the credibility of the detected target object is higher, thereby improving the accuracy of the front-end analysis device in detecting the target event, that is, improving the positive reporting rate of the target event reported by the front-end analysis device.

[0142] In the case where the object size threshold needs to be reduced, the adjustment condition of the object size threshold includes that the minimum size is less than the object size threshold; or, in the case where the object size threshold needs to be expanded, the adjustment condition of the object size threshold includes that the minimum size is greater than the object size threshold. The minimum size is the minimum value of the size of the target object obtained after the center end recognizes the original image associated with the positive analysis result.

[0143] For example, if during the adjustment period, the center determines that the minimum size of the target object in the recognition result of the positive event match is 30×15, assuming that the adjustment of the target representation requires reducing the object size threshold, and the original object size threshold is 40×20, then the minimum size is smaller than the original object size threshold, indicating that the object size threshold meets the adjustment conditions and the object size threshold can be reduced. Assuming that the adjustment of the target representation requires expanding the object size threshold, and the original object size threshold is 28×15, then the minimum size is larger than the original object size threshold, indicating that the object size threshold meets the adjustment conditions and the object size threshold can be expanded.

[0144] When specifically reducing or expanding the object size threshold, the object size threshold may be reduced or expanded by a preset ratio, or the object size threshold may be reduced / expanded by a reduction variable or an expansion variable corresponding to the difference between the minimum reported size and the object size threshold, ... .

[0145] In this way, by adjusting the object size threshold of the target objects filtered and reported by the front-end analysis device, the target objects missed by the front-end analysis device can be reduced or increased, and the detection rate of the front-end analysis device can be adjusted.

[0146] In the case where the detection sensitivity needs to be reduced, the adjustment condition of the detection sensitivity includes that the third ratio is less than the preset ratio; or, in the case where the detection sensitivity needs to be increased, the adjustment condition of the detection sensitivity includes that the third ratio is greater than or equal to the preset ratio. The third ratio is used to characterize the correct rate of the target object intersecting with the boundary of the detection area of ​​the front-end analysis device in the target event reported by the front-end analysis device, and the third ratio is obtained based on the positive analysis result and the wrong analysis result.

[0147] Specifically, the electronic device may determine the third ratio based on the number of target objects intersecting the boundary of the detection area of ​​the front-end analysis device in all positive analysis results and the number of target objects intersecting the boundary of the detection area of ​​the front-end analysis device in the false analysis results.

[0148] For example, assuming that within the adjustment period, the front-end analysis device reports 15 target events, of which 12 correspond to positive analysis results and 3 correspond to false analysis results. The number of target objects that intersect with the boundary of the detection area in the 12 positive analysis results is 25, and the number of target objects that intersect with the boundary of the detection area in the 3 false analysis results is 5. Then the third ratio is 25÷(25+5)=83.3%. Assuming that the detection sensitivity needs to be reduced and the preset ratio is 85%, the third ratio is less than the preset ratio, and the detection sensitivity can be reduced. Assuming that the detection sensitivity needs to be increased and the preset ratio is 80%, the third ratio is greater than the preset ratio, and the detection sensitivity can be increased.

[0149] When actually reducing or increasing the detection sensitivity, the detection sensitivity may be reduced / increased according to a preset ratio, or the detection sensitivity of a fixed variable may be reduced / increased each time, or the detection sensitivity may be reduced / increased according to the reduced / increased variable corresponding to the difference between the third ratio and the preset ratio, etc. For example, if the detection sensitivity is reduced by 2%, the detection sensitivity before the reduction is c, and the detection sensitivity after the reduction is 98%c.

[0150] In this way, by adjusting the detection sensitivity of the front-end analysis device, it is possible to avoid too many false alarm events caused by too high detection sensitivity, or to increase the number of target events detected by the front-end analysis device. Therefore, the detection rate of the front-end analysis device for the target events can be adjusted by adjusting the detection sensitivity.

[0151] When the reporting interval threshold needs to be shortened, the adjustment conditions of the reporting interval threshold include that the reporting interval threshold is greater than the shortest duration; or, when the reporting interval threshold needs to be increased, the adjustment conditions of the duration threshold include that the reporting interval threshold is less than the shortest duration; the shortest duration is the shortest value of the duration of the target event obtained after the center end identifies the original image associated with the positive analysis result.

[0152] For example, for a video file within a period of time, both the front-end analysis device and the central end identify target events for the video file. If, in the positive analysis results corresponding to the target events identified by the front-end analysis device, the shortest duration of the target event is 10s, and the reporting interval threshold of the front-end analysis device is 15s, then the preset duration of the front-end analysis device should be shortened to detect target events with shorter durations to avoid missing target events.

[0153] In this way, by adjusting the reporting interval threshold of the front-end analysis device, the detection effect of the front-end analysis device can be made closer to the detection effect of the central end.

[0154] Before formally identifying the target event, you can set an optimization cycle to preliminarily optimize the identification parameters of the front-end analysis device. During the optimization cycle, the front-end analysis device and the center both perform the same event detection process on the same video file. Such events contain the attributes of continuous events, such as falling to the ground, playing with mobile phones, etc. The effective duration of all missed events during the optimization cycle is counted, and a certain coefficient is set. If the missed events with a duration less than the reporting interval threshold of the front-end analysis device reach the threshold coefficient, the reporting interval threshold of the front-end analysis device can be set by linearly expanding with a certain coefficient based on the average duration of the target event in the missed events or by expanding based on certain fixed values.

[0155] The setting of the reporting interval threshold is related to the specific installation scene of the image acquisition device, the target event type, the detection time period and other factors. For example, the image acquisition device is set up in a commercial street, and the target event is to detect people who fall down. Since the flow of people in the commercial street is large, if a person falls down, the possibility of the person being quickly helped up is very high. If the reporting interval threshold is long, the person who is quickly helped up may not be detected. For another example, the image acquisition device is set up on a remote path, and the target event is to detect people who fall down. The flow of people is small. If a person falls down, he may not be helped up for a long time. In this case, the reporting interval threshold can be set longer to effectively detect the person who falls down. For another example, the target event is a fight, and multiple people conflict and cause a fight. If the reporting interval threshold is short, the target event may be missed. For another example, the flow of people at the same location in different time periods (such as daytime, night, weekdays, holidays, etc.) is different, then the target event for personnel detection will have different detection effects, and the reporting interval threshold can be set for different time periods.

[0156] In some embodiments, the adjustable identification parameter, the adjustment direction of the adjustable identification parameter, and the adjustment condition of the adjustable identification parameter in the parameter adjustment strategy corresponding to the adjustment target may be obtained by performing multiple tests on the electronic device.

[0157] S103, analyzing the adjustable identification parameters in the analysis results of each target event in the multiple target events, and when the analysis results of the adjustable identification parameters meet the adjustment conditions, controlling the front-end analysis device to update the adjustable identification parameters according to the adjustment direction.

[0158] In some embodiments, when the adjustment target is a single target, the electronic device controls the front-end analysis device to update the adjustable identification parameter according to the adjustment direction, which can be specifically implemented as follows: the electronic device determines the first weight of the adjustable identification parameter in the adjustment direction based on the degree of influence of each adjustable identification parameter on the adjustment target in the adjustment direction; the electronic device determines the adjustment amplitude of the adjustable identification parameter based on the first weight of the adjustable identification parameter; the electronic device controls the front-end analysis device to update the adjustable identification parameter according to the adjustment direction and the adjustment amplitude.

[0159] The first weight and the adjustment range may be positively correlated, and the larger the first weight is, the larger the adjustment range is.

[0160] For example, if the current computing power utilization rate of the central end is 70%, and the adjustment target represents that the user expects the computing power utilization rate of the central end to be reduced to 60%, then it can be determined that the adjustable recognition parameters include the detection area and the shielding area, the adjustment direction of the detection area is to reduce, and the first weight of the detection area is 80%, and the adjustment direction of the shielding area is to expand, and the first weight of the shielding area is 20%, then the adjustment range of the detection area is greater than the adjustment range of the shielding area. Assuming that the adjustment range of the detection area is to shorten the side length of the detection area by 20%, and the adjustment range of the shielding area is to increase the side length of the shielding area by 5%, if the resolution of the original detection area is a×b, then the resolution of the reduced detection area is 80%a×80%b, and if the resolution of the original shielding area is c×d, then the resolution of the expanded shielding area is 105%c×105%d.

[0161] In this way, the electronic device can control the front-end analysis device to update the adjustable recognition parameters according to the adjustment direction and adjustment amplitude. After the parameters are updated, the detection effect of the front-end analysis device on the target event is closer to the detection effect of the central end.

[0162] Optionally, the electronic device may determine the number of times each adjustable parameter is adjusted based on the first weight of each adjustable identification parameter, and then adjust the adjustable identification parameter according to the adjustment direction based on the preset adjustment amplitude and the number of adjustments corresponding to the first weight. For example, the preset adjustment amplitude is 1, and taking the example of improving the positive reporting rate of the front-end analysis device for the target event as an example, it is necessary to increase the confidence threshold, and the confidence threshold may be increased by 1 each time the confidence threshold is adjusted, and multiple adjustments are made according to the preset number of adjustments until the positive reporting rate requirement of the front-end analysis device for detecting the target event is met.

[0163] In some embodiments, when the adjustment target is multiple targets, the electronic device controls the front-end analysis device to update the adjustable identification parameters according to the adjustment direction, which can be specifically implemented as follows: the electronic device determines the second weight of each single target in the adjustment target, and the second weight is used to characterize the importance of the single target to the adjustment target; the electronic device determines the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and determines the adjustment amplitude of the adjustable identification parameter; the electronic device determines the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target based on the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and the second weight of each single target in the adjustment target; the electronic device determines the adjustment frequency of each adjustable identification parameter involved in the adjustment target based on the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target; the electronic device controls the front-end analysis device to update the adjustable identification parameters according to the adjustment direction, adjustment amplitude, and adjustment frequency.

[0164] In many business scenarios, it is necessary to make a comprehensive optimal decision between the computing power utilization rate at the center end, the detection rate of the target event by the front-end analysis equipment, and the positive reporting rate of the target event by the front-end analysis equipment. When doing a comprehensive analysis, make a weight agreement for each target in the adjustment target (which can be set based on experience). For example, set the computing power utilization rate at the center end to correspond to weight A, the detection rate of the target event by the front-end analysis equipment to correspond to weight B, and the positive reporting rate of the target event by the front-end analysis equipment to correspond to weight C. Assuming that only the three adjustable recognition parameters of detection area, shielding area and confidence threshold are adjusted, the highest impact factor is 10, the impact factor of the detection area on the computing power utilization rate of the center is 7, the impact factor of the shielding area on the computing power utilization rate of the center is 3, and the impact factor of the confidence threshold on the computing power utilization rate of the center is 9; the impact factor of the detection area on the detection rate of the target event by the front-end analysis device is 5, the impact factor of the shielding area on the detection rate of the target event by the front-end analysis device is 4, and the impact factor of the confidence threshold on the detection rate of the target event by the front-end analysis device is 8; the impact factor of the detection area on the positive alarm rate of the target event by the front-end analysis device is 8, the impact factor of the shielding area on the positive alarm rate of the target event by the front-end analysis device is 5, and the impact factor of the confidence threshold on the positive alarm rate of the target event by the front-end analysis device is 6. Then the influence factor of the detection area on the comprehensive result of the adjustment target is (7A+5B+8C); the influence factor of the shielding area on the comprehensive result of the adjustment target is (3A+4B+5C); the influence factor of the confidence threshold on the comprehensive result of the adjustment target is (9A+8B+6C); finally, the adjustment frequency of the detection area, shielding area, and confidence threshold is carried out in the ratio of (7A+5B+8C): (3A+4B+5C): (9A+8B+6C). Assuming A is 50%, B is 30%, and C is 20%, the ratio is 6.6:3.7:8.1=66:37:81, indicating that more adjustments are used to adjust the detection area and the confidence threshold. Then, the electronic device determines the adjustment frequency of the detection area, shielding area, and confidence threshold based on this ratio, and then controls the front-end analysis device to update the adjustable recognition parameters based on the adjustment direction and adjustment amplitude corresponding to these three adjustable recognition parameters.

[0165] In this way, the update of parameters fully considers the impact of each recognition parameter on the adjustment target. After the parameter update, the detection effect of the front-end analysis device on the target event is closer to the detection effect of the center end.

[0166] In the process of actually adjusting the adjustable recognition parameters of the front-end analysis device, multiple adjustments can be made to achieve the adjustment target. For example, if the adjustment target is to reduce the computing power utilization rate of the central end to 70%, after adjusting the adjustable recognition based on the adjustment strategy corresponding to the adjustment target, if the computing power utilization rate of the central end has not actually dropped to 70%, the adjustable recognition parameters can be adjusted again, and the adjustment target can be finally met through multiple adjustments.

[0167] Figure 3 The technical solution shown brings at least the following beneficial effects: after the central end obtains the target event reported by the front-end analysis device, it performs a true / false alarm analysis to obtain the analysis result of each target event. The present application determines the adjustment target of parameter optimization and the parameter adjustment strategy corresponding to the adjustment target, that is, it determines the effect to be achieved by adjusting the front-end analysis device and the central end to collaboratively identify the target event, and how to adjust the parameters of the front-end analysis device when identifying the target event. Finally, based on the adjustable identification parameters in the analysis results of these target events, it is determined whether to adjust the adjustable identification parameters and control the front-end analysis device to adjust the adjustable identification parameters. The entire parameter adjustment process is performed by the device itself to adjust the parameters when the front-end analysis device identifies the target event, and the user does not need to manually adjust the parameters, which can save the user's operation steps. In addition, the adjustable identification parameters are the identification parameters when the front-end analysis device identifies the target event. The adjustment target represents the user's needs. The effect of the front-end analysis device and the central end collaboratively detecting the target event after adjustment meets the adjustment target, which also meets the user's requirements for the front-end analysis device and the central end to collaboratively detect the target event. At the same time, the front-end analysis device configured by the user is also put into use, and the detection capability of the front-end analysis device is not wasted.

[0168] The following describes the solution of this application from the perspective of the overall process:

[0169] Take the electronic device as the central end as an example. Fig.10As shown, in the process of adjusting the recognition parameters of the front-end analysis device, the central end receives the target event reported by the front-end analysis device and the original image associated with the target event, the central end recognizes the original image, obtains the recognition result, and determines the true / false alarm analysis result of the target event reported by the front-end analysis device based on the recognition result, that is, determines whether the target event reported by the front-end analysis device is a true alarm event or a false alarm event (in this process, manual verification can be performed to verify whether the true / false alarm analysis result determined by the central end is accurate), and then the central end determines each recognition parameter (detection area, shielding area, confidence threshold, object size threshold, detection sensitivity, and reporting interval threshold) based on the true / false alarm analysis result, and then adjusts the adjustable recognition parameters according to the adjustment direction and adjustment conditions of the adjustable recognition parameters in the parameter adjustment strategy corresponding to the adjustment target in combination with the adjustment target. This adjustment process is iterated many times, and finally makes the detection effect of the front-end analysis device closer to the detection effect of the central end.

[0170] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to achieve the above functions, it includes hardware structures and / or software modules corresponding to the execution of each function. It should be easy to realize that the technical goals in this field are combined with the units and algorithm steps of each example described in the embodiments disclosed in this article, and the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical goals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0171] like Fig.11 As shown, the embodiment of the present application also provides a parameter optimization device for target identification, which is used to execute the parameter optimization method for target identification shown in the above method embodiment. The parameter optimization device 500 for target identification includes: a transmission module 501 and a processing module 502.

[0172] Among them, the transmission module 501 is used to obtain the analysis result of each target event among multiple target events; the multiple target events are reported to the central end by the front-end analysis device; the analysis result is obtained by the central end by performing a true / false alarm analysis on the multiple target events, and the analysis result is used to indicate whether the target event is a false alarm; the processing module 502 is used to determine the adjustment target of the parameter optimization; the processing module 502 is also used to determine the parameter adjustment strategy corresponding to the adjustment target, and the parameter adjustment strategy includes the adjustment conditions and adjustment directions of the adjustable identification parameters; the adjustable identification parameters are the identification parameters when the front-end analysis device identifies the target events; the processing module 502 is also used to analyze the adjustable identification parameters in the analysis results of each target event among the multiple target events, and when the analysis results of the adjustable identification parameters meet the adjustment conditions, control the front-end analysis device to update the adjustable identification parameters according to the adjustment direction.

[0173] In one possible implementation, the adjustment target includes reducing or increasing the computing power utilization rate of the central end; when the adjustment target includes reducing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area and / or shielding area when the front-end analysis device identifies the target object; the adjustment direction of the detection area includes reducing the detection area, and the adjustment direction of the shielding area includes expanding the shielding area; when the adjustment target includes increasing the computing power utilization rate of the central end, the adjustable recognition parameters in the parameter adjustment strategy include the detection area, and the adjustment direction of the detection area includes expanding the detection area.

[0174] In another possible implementation, the adjustment target includes lowering or increasing the detection rate of the target event by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the filtering parameters of the target event by the front-end analysis device; the filtering parameters include at least one of the confidence threshold, object size threshold, detection sensitivity and reporting interval threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the detection rate, the adjustment direction of the filtering parameters includes at least one of reducing the confidence threshold, reducing the object size threshold, reducing the detection sensitivity and shortening the reporting interval threshold; when the adjustment target includes reducing the detection rate, the adjustment direction of the filtering parameters includes at least one of increasing the confidence threshold, expanding the object size threshold, improving the detection sensitivity and increasing the reporting interval threshold.

[0175] In another possible implementation, the adjustment target includes lowering or increasing the positive reporting rate of target events reported by the front-end analysis device; the adjustable identification parameters in the parameter adjustment strategy include the confidence threshold when the front-end analysis device identifies the target object; when the adjustment target includes increasing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to increase the confidence threshold; when the adjustment target includes reducing the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to reduce the confidence threshold.

[0176] In another possible implementation, when the detection area needs to be reduced, the adjustment condition of the detection area includes any one of the following: the reference area is smaller than the detection area, the first area is smaller than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference, the first ratio is smaller than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein, when the adjustment condition of the detection area includes the case where the reference area is smaller than the detection area, the adjustment direction of the detection area is to reduce the detection area in a direction close to the reference area; or, when the detection area needs to be expanded, the adjustment condition of the detection area includes any one of the following: the first area is larger than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to the first preset difference. Assume that the difference, the first ratio is greater than the second ratio, and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to the second preset difference; wherein the analysis result is a positive analysis result or an erroneous analysis result; the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result, the first area is the area occupied by the target object in the original image obtained after the center end performs target event recognition on the original image associated with the positive analysis result, the second area is the area occupied by the target object in the positive analysis result in the detection area of ​​the front-end analysis device, the first ratio is the ratio between the first area and the area of ​​the detection area when the front-end analysis device performs target event recognition, and the second ratio is the ratio between the second area and the area of ​​the detection area when the front-end analysis device performs target event recognition.

[0177] In another possible implementation, when it is necessary to expand the shielding area, the adjustment conditions of the shielding area include the existence of a false alarm area, and the expansion of the shielding area is specifically implemented by merging the false alarm area into the original shielding area to expand the shielding area, and the false alarm area is obtained based on the positive analysis result in the analysis result of the target event and the position of the target object in the erroneous analysis result on the original image associated with the positive analysis result and the erroneous analysis result.

[0178] In another possible implementation, when the object size threshold needs to be reduced, the adjustment condition of the object size threshold includes that the minimum size is less than the object size threshold; or, when the object size threshold needs to be expanded, the adjustment condition of the object size threshold includes that the minimum size is greater than the object size threshold; or, when the detection sensitivity needs to be reduced, the adjustment condition of the detection sensitivity includes that the third ratio is less than the preset ratio; or, when the detection sensitivity needs to be increased, the adjustment condition of the detection sensitivity includes that the third ratio is greater than or equal to the preset ratio; or, when the reporting interval threshold needs to be shortened, the adjustment condition of the reporting interval threshold includes that the reporting interval threshold is greater than the shortest duration. continuation duration; or, in the case where it is necessary to increase the reporting interval threshold, the adjustment condition of the continuation duration threshold includes that the reporting interval threshold is less than the shortest continuation duration; wherein the analysis result is a positive analysis result or an erroneous analysis result; the minimum size is the minimum value of the size of the target object obtained after the center end identifies the original image associated with the positive analysis result; the third ratio is used to characterize the accuracy of the target object that intersects with the boundary of the detection area of ​​the front-end analysis device in the target event reported by the front-end analysis device, and the third ratio is obtained based on the positive analysis result and the erroneous analysis result; the shortest continuation duration is the shortest value of the continuation duration of the target event obtained after the center end identifies the original image associated with the positive analysis result.

[0179] In another possible implementation, when the confidence threshold needs to be reduced, the adjustment condition of the confidence threshold includes that the false alarm rate is greater than or equal to a preset false alarm rate; or, when the confidence threshold needs to be increased, the adjustment condition of the confidence threshold includes that the false alarm rate is less than a preset false alarm rate; wherein the analysis result is a positive analysis result or an erroneous analysis result; and the false alarm rate is the ratio between the number of erroneous analysis results in the analysis results of multiple target events and the number of all analysis results.

[0180] In another possible implementation, the processing module 502 is specifically used to: when the adjustment target is a single target, determine the first weight of the adjustable identification parameter in the adjustment direction based on the degree of influence of each adjustable identification parameter on the adjustment target in the adjustment direction; determine the adjustment range of the adjustable identification parameter based on the first weight of the adjustable identification parameter; and control the front-end analysis device to update the adjustable identification parameter according to the adjustment direction and adjustment range.

[0181] In another possible implementation, the processing module 502 is specifically used to: when the adjustment target is a multi-target, determine the second weight of each single target in the adjustment target, the second weight is used to characterize the importance of the single target to the adjustment target; determine the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and determine the adjustment range of the adjustable identification parameter; based on the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and the second weight of each single target in the adjustment target, determine the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target; based on the comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target, determine the adjustment frequency of each adjustable identification parameter involved in the adjustment target; control the front-end analysis device to update the adjustable identification parameter according to the adjustment direction, adjustment range, and adjustment frequency.

[0182] It should be noted that Fig.11 The division of modules in the example is schematic and is only a logical function division. There may be other division methods in actual implementation. For example, two or more functions may be integrated into one processing module. The above integrated modules may be implemented in the form of hardware or software function modules.

[0183] Another embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the parameter optimization method for target identification shown in the above embodiment.

[0184] In actual implementation, the transmission module 501 and the processing module 502 can be implemented by the processor of the electronic device calling the computer program code in the memory. The specific execution process can refer to the description of the parameter optimization method of the target identification above, which will not be repeated here.

[0185] Another embodiment of the present application further provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed by a processor, the steps of the parameter optimization method for target identification shown in the above embodiment are implemented.

[0186] In another embodiment of the present application, a computer program product is provided. The computer program product includes computer instructions. When the computer instructions are executed by a processor, the steps of the parameter optimization method for target identification shown in the above embodiment are implemented.

[0187] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using a software program, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer execution instruction is loaded and executed on the computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instruction can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instruction can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that can be accessed by the computer or a data storage device such as a server, data center, etc. that contains one or more servers that can be integrated with the medium. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), etc.

[0188] The above is only a specific implementation of the present application. Those skilled in the art may conceive of changes or substitutions based on the specific implementation provided by the present application, which should all be included in the protection scope of the present application.

Claims

1. A parameter optimization method for target recognition, characterized in that: include: Obtaining analysis results for each target event among multiple target events; The multiple target events are obtained by the front-end analysis device through identifying the image and reported to the central end; The analysis result is obtained by the central end performing a true / false alarm analysis on the multiple target events, and the analysis result is used to indicate whether the target event is a false alarm. Whether the target event is a false alarm is determined based on the recognition result obtained by the central end performing target event recognition on the original image associated with the target event reported by the front-end analysis device, and whether it matches the target event reported by the front-end analysis device; Determine an adjustment target for parameter optimization, wherein the adjustment target includes one or more of the following: reducing or increasing the utilization rate of computing power at the center end, reducing or increasing the detection rate of target events by the front-end analysis device, and reducing or increasing the positive reporting rate of target events reported by the front-end analysis device; Determine a parameter adjustment strategy corresponding to the adjustment target, wherein the parameter adjustment strategy includes an adjustment condition and an adjustment direction of an adjustable identification parameter; the adjustable identification parameter is an identification parameter when the front-end analysis device performs target event identification; The adjustable identification parameter in the analysis result of each target event in the multiple target events is analyzed, and when the analysis result of the adjustable identification parameter meets the adjustment condition, the front-end analysis device is controlled to update the adjustable identification parameter according to the adjustment direction.

2. The method according to claim 1, characterized in that: The adjustment target includes reducing or increasing the computing power utilization rate of the central end; In the case where the adjustment target includes reducing the utilization rate of the computing power of the central end, the adjustable identification parameter in the parameter adjustment strategy includes the detection area and / or the shielding area when the front-end analysis device performs target object identification; the adjustment direction of the detection area includes reducing the detection area, and the adjustment direction of the shielding area includes expanding the shielding area; In the case where the adjustment target includes improving the utilization rate of the computing power of the central end, the adjustable identification parameter in the parameter adjustment strategy includes the detection area, and the adjustment direction of the detection area includes expanding the detection area.

3. The method according to claim 1, characterized in that The adjustment target includes reducing or increasing the detection rate of the front-end analysis device for the target event; The adjustable identification parameters in the parameter adjustment strategy include filtering parameters of the front-end analysis device for target events; the filtering parameters include at least one of a confidence threshold, an object size threshold, a detection sensitivity, and a reporting interval threshold when the front-end analysis device identifies a target object; In the case where the adjustment target includes improving the detection rate, the adjustment direction of the filtering parameter includes at least one of reducing the confidence threshold, reducing the object size threshold, reducing the detection sensitivity, and shortening the reporting interval threshold; In the case where the adjustment target includes reducing the detection rate, the adjustment direction of the filtering parameter includes at least one of increasing the confidence threshold, expanding the object size threshold, improving the detection sensitivity, and increasing the reporting interval threshold.

4. The method according to claim 1, characterized in that: The adjustment target includes reducing or increasing the positive reporting rate of the target event reported by the front-end analysis device; the adjustable identification parameter in the parameter adjustment strategy includes the confidence threshold when the front-end analysis device identifies the target object; In the case where the adjustment target includes improving the positive reporting rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to increase the confidence threshold; In the case where the adjustment target includes reducing the positive alarm rate, the adjustment direction of the confidence threshold in the parameter adjustment strategy is to reduce the confidence threshold.

5. The method according to claim 2, characterized in that: In the case where the detection area needs to be reduced, the adjustment condition of the detection area includes any one of the following: the reference area is smaller than the detection area, the first area is smaller than the second area and the absolute value of the difference between the first area and the second area is greater than or equal to a first preset difference, the first ratio is smaller than the second ratio and the absolute value of the difference between the first ratio and the second ratio is greater than or equal to a second preset difference; wherein, in the case where the adjustment condition of the detection area includes that the reference area is smaller than the detection area, the adjustment direction of the detection area is to reduce the detection area in a direction close to the reference area; or, In the case where the detection area needs to be expanded, the adjustment condition of the detection area includes any one of the following: the first area is larger than the second area and the absolute value of the difference between the first area and the second area is larger than or equal to a first preset difference, the first ratio is larger than the second ratio and the absolute value of the difference between the first ratio and the second ratio is larger than or equal to a second preset difference; Wherein, the analysis result is a positive analysis result or an erroneous analysis result; the reference area is obtained based on the position of the target object in the positive analysis result on the original image associated with the positive analysis result; the first area is the area occupied by the target object in the original image obtained after the center end performs target event recognition on the original image associated with the positive analysis result; the second area is the area occupied by the target object in the positive analysis result in the detection area of ​​the front-end analysis device; the first ratio is the ratio between the first area and the area of ​​the detection area when the front-end analysis device performs target event recognition; the second ratio is the ratio between the second area and the area of ​​the detection area when the front-end analysis device performs target event recognition.

6. The method according to claim 2, characterized in that In the case where it is necessary to expand the shielding area, the adjustment conditions of the shielding area include the existence of a false alarm area, and the expansion of the shielding area is specifically implemented by merging the false alarm area into the original shielding area to expand the shielding area. The false alarm area is obtained based on the positions of the target objects in the positive analysis results and the false analysis results in the analysis results of the target event on the original image associated with the positive analysis results and the false analysis results.

7. The method according to claim 3, characterized in that In the case where the object size threshold needs to be reduced, the adjustment condition of the object size threshold includes that the minimum size is smaller than the object size threshold; or, In the case where the object size threshold needs to be enlarged, the adjustment condition of the object size threshold includes that the minimum size is greater than the object size threshold; or, In the case where the detection sensitivity needs to be reduced, the adjustment condition of the detection sensitivity includes that the third ratio is less than a preset ratio; or, In the case where the detection sensitivity needs to be improved, the adjustment condition of the detection sensitivity includes that the third ratio is greater than or equal to the preset ratio; or, In the case where the reporting interval threshold needs to be shortened, the adjustment condition of the reporting interval threshold includes that the reporting interval threshold is greater than the shortest duration; or, In the case where the reporting interval threshold needs to be increased, the adjustment condition of the duration threshold includes that the reporting interval threshold is less than the shortest duration; The analysis result is a positive analysis result or an erroneous analysis result; the minimum size is the minimum value of the size of the target object obtained after the center end recognizes the original image associated with the positive analysis result; The third ratio is used to characterize the accuracy of the target objects that intersect with the boundary of the detection area of ​​the front-end analysis device in the target events reported by the front-end analysis device, and the third ratio is obtained based on the positive analysis result and the incorrect analysis result; the shortest duration is the shortest value of the duration of the target event obtained after the central end identifies the original image associated with the positive analysis result.

8. The method according to claim 4, characterized in that In the case where the confidence threshold needs to be reduced, the adjustment condition of the confidence threshold includes that the false alarm rate is greater than or equal to the preset false alarm rate; or, In the case where the confidence threshold needs to be increased, the adjustment condition of the confidence threshold includes that the false alarm rate is less than a preset false alarm rate; The analysis result is a positive analysis result or an erroneous analysis result; and the false alarm rate is a ratio between the number of erroneous analysis results and the number of all analysis results in the analysis results of the multiple target events.

9. The method according to any one of claims 1 to 8, characterized in that: The controlling the front-end analysis device to update the adjustable identification parameter according to the adjustment direction includes: In the case where the adjustment target is a single target, determining a first weight of the adjustable identification parameter in the adjustment direction based on the influence degree of each adjustable identification parameter on the adjustment target in the adjustment direction; Determining an adjustment range of the adjustable identification parameter based on a first weight of the adjustable identification parameter; The front-end analysis device is controlled to update the adjustable identification parameter according to the adjustment direction and the adjustment amplitude.

10. The method according to any one of claims 1 to 8, characterized in that: The controlling the front-end analysis device to update the adjustable identification parameter according to the adjustment direction includes: In the case where the adjustment target is a multi-target, determining a second weight of each single target in the adjustment target, wherein the second weight is used to represent the importance of the single target to the adjustment target; Determine the influence factor of each adjustable identification parameter corresponding to each single target on the single target, and determine the adjustment range of the adjustable identification parameter; Determine a comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target based on the influence factor of each adjustable identification parameter corresponding to each single target on the single target and the second weight of each single target in the adjustment target; Determining an adjustment frequency of each adjustable identification parameter involved in the adjustment target based on a comprehensive influence factor of each adjustable identification parameter involved in the adjustment target on the adjustment target; The front-end analysis device is controlled to update the adjustable identification parameter according to the adjustment direction, the adjustment amplitude, and the adjustment frequency.

11. A parameter optimization device for target recognition, characterized in that: include: A transmission module, used for obtaining analysis results of each target event among a plurality of target events; The multiple target events are obtained by the front-end analysis device through identifying the image and reported to the central end; The analysis result is obtained by the central end performing a true / false alarm analysis on the multiple target events, and the analysis result is used to indicate whether the target event is a false alarm. Whether the target event is a false alarm is determined based on the recognition result obtained by the central end performing target event recognition on the original image associated with the target event reported by the front-end analysis device, and whether it matches the target event reported by the front-end analysis device; A processing module, used to determine an adjustment target for parameter optimization, wherein the adjustment target includes one or more of the following: reducing or increasing the utilization rate of the computing power of the central end, reducing or increasing the detection rate of the target event by the front-end analysis device, and reducing or increasing the positive reporting rate of the target event reported by the front-end analysis device; The processing module is also used to determine the parameter adjustment strategy corresponding to the adjustment target, wherein the parameter adjustment strategy includes the adjustment conditions and adjustment directions of the adjustable identification parameters; the adjustable identification parameters are the identification parameters when the front-end analysis device performs target event identification; The processing module is also used to analyze the adjustable identification parameter in the analysis result of each target event in the multiple target events, and when the analysis result of the adjustable identification parameter meets the adjustment condition, control the front-end analysis device to update the adjustable identification parameter according to the adjustment direction.

12. A target recognition system, characterized in that: Including electronic equipment, front-end analysis equipment and central end; The front-end analysis device is used to detect target events in the image of the video file, and report the detected multiple target events to the central end; The central end is used to receive the multiple target events reported by the front-end analysis device, and perform true / false positive analysis on the multiple target events to obtain an analysis result of each target event in the multiple target events; The analysis result is used to indicate whether the target event is a false alarm, and whether the target event is a false alarm is determined based on the recognition result obtained by the center end for the target event recognition of the original image associated with the target event reported by the front-end analysis device, and whether it matches the target event reported by the front-end analysis device; The electronic device is used to execute the parameter optimization method for target identification as described in any one of claims 1-10.

13. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 10.

Images