Target behavior detection method and device, storage medium and electronic device

Abstract

The application provides a target behavior detection method and device, a storage medium and an electronic device. The target behavior detection method comprises the following steps: setting a false detection filtering template according to a plurality of historical monitoring images; identifying and removing a false detection object in a current monitoring image through the false detection filtering template to determine a first detection box of a target object in the current monitoring image; and determining that a target behavior exists in the case that the first detection box is in a target region in the current monitoring image.

Description

Technical Field

[0001] This application relates to the field of intelligent security technology, and more specifically, to a method, apparatus, storage medium, and electronic device for detecting target behavior. Background Technology

[0002] In related technologies, the detection scheme for violations first identifies the target object in the monitoring image, then determines whether the target object is in the violation area of ​​the monitoring image, and judges whether there is a violation. When using this scheme, the complex background of the monitoring image may lead to false detection of the target object, resulting in false reports of violations. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the related art.

[0004] Therefore, the first aspect of this application is to propose a method for detecting target behavior.

[0005] The second aspect of this application is to propose a device for detecting target behavior.

[0006] The third aspect of this application is to propose another detection device for target behavior.

[0007] The fourth aspect of this application is to propose a readable storage medium.

[0008] The fifth aspect of this application is to propose an electronic device.

[0009] The sixth aspect of this application is to provide a computer program product.

[0010] In view of this, according to one aspect of this application, a method for detecting target behavior is proposed, the method comprising: setting a false detection filtering template based on multiple historical surveillance images; identifying and removing falsely detected objects in the current surveillance image through the false detection filtering template to determine a first detection box of the target object in the current surveillance image; and determining the existence of target behavior if the first detection box is located in the target area of ​​the current surveillance image.

[0011] It should be noted that the execution subject of the target behavior detection method proposed in this application can be a target behavior detection device. In order to more clearly explain the target behavior detection method proposed in this application, the following technical solution uses the target behavior detection device as the execution subject of the target behavior detection method for illustrative purposes.

[0012] In this technical solution, the aforementioned false detection objects refer to objects in the monitoring image that are easily identified as target objects but are not target objects, such as scarecrows or robots in a park; the aforementioned historical monitoring images refer to historical images captured from the monitoring video of the monitoring area; the aforementioned current monitoring image refers to the current image captured from the monitoring video of the monitoring area; the aforementioned false detection filtering template refers to a filtering template for false detection objects that are continuously falsely detected as target objects in the background of the monitoring image; the aforementioned first detection box refers to the frame line of the target object identified in the current monitoring image; the aforementioned target area refers to the violation area in the current monitoring image, such as a non-zebra crossing area; the aforementioned target behavior refers to the violation behavior, such as pedestrian violation behavior.

[0013] Specifically, the detection device first determines a false detection filtering template for that portion of the surveillance images based on multiple historical surveillance images. More specifically, based on these historical images, it can identify falsely detected objects in that portion of the surveillance images that are easily identifiable as the target object. When identifying the target object in a new portion of the surveillance images, the new images can be processed based on these falsely detected objects. Therefore, the detection device is able to determine the false detection filtering template based on multiple historical surveillance images.

[0014] Furthermore, the detection device identifies the first detection box of the target object in the current monitoring image. Specifically, the detection device first identifies the target object in the current monitoring image, then removes falsely detected objects that were mistakenly identified as the target object from the identified target objects according to the aforementioned false detection filtering template, and finally determines the detection boxes of the remaining target objects, namely the aforementioned first detection boxes.

[0015] Furthermore, the detection device determines whether the first detection frame is located within the target area set in the current monitoring image. Specifically, if it is determined that the first detection frame is located within the target area, the detection device determines that there is a target behavior in the monitoring image.

[0016] Specifically, if the first detection box is located in the target area, it indicates that the target object is in a violation area. Therefore, the detection device can determine that there is target behavior in the monitoring image.

[0017] For example, the process of this application is explained by taking the false detection object as an object resembling a pedestrian, the target object as a pedestrian, the target area as a non-zebra crossing area, and the target behavior as a pedestrian violation.

[0018] Specifically, the detection device first identifies objects that are easily misidentified as pedestrians based on historical surveillance images near the zebra crossing, and then sets up false detection filtering templates based on these objects.

[0019] Furthermore, the detection device identifies pedestrians in the current surveillance image near the zebra crossing and removes non-pedestrian objects from the identified pedestrians according to the false detection filtering template, so as to determine the first detection box for all pedestrians in the current surveillance image.

[0020] Furthermore, the detection device determines whether the pedestrian's first detection frame is located in a non-zebra crossing area in the current monitoring image. If it is located in a non-zebra crossing area, it is determined that there is a pedestrian violation in the current monitoring image.

[0021] For example, the target behavior detection method proposed in this application can be used in intelligent security scenarios in parks or factories. Specifically, the target behaviors included in parks or factories mainly include production line crossing behavior, park intrusion behavior, pedestrian violation behavior, and vehicle illegal parking behavior.

[0022] In this technical solution, the detection device can determine a false detection filtering template for a portion of the monitoring images based on multiple historical monitoring images. This template removes falsely identified objects that are easily mistaken for target objects during the identification process in the current monitoring image. A first detection box is then established for all target objects. Based on the determination of whether the first detection box is within the target area, the presence of target behavior can be determined. In this application's technical solution, the detection device can filter out falsely identified objects using the set false detection filtering template, solving the problem in related technologies where complex backgrounds in monitoring images lead to false detections of target objects and false alarms of violations, thus ensuring the accuracy of target behavior detection.

[0023] Furthermore, the target behavior detection method proposed according to the above-described technical solution of the present invention may also have the following additional technical features:

[0024] In the above technical solution, the step of setting a false detection filtering template based on multiple historical surveillance images specifically includes: identifying a second detection box in multiple historical surveillance images and determining a first score and a first coordinate of the second detection box; determining the detection box of the falsely detected object in the second detection box based on the first score and the first coordinate; and setting a false detection filtering template based on the detection box of the falsely detected object; wherein, the second detection box is the detection box of the falsely detected object and the target object in multiple historical surveillance images, and the first score is the probability score of the second detection box being the first detection box.

[0025] In this technical solution, the detection device can determine the detection boxes of falsely detected objects included in the second detection box based on the first coordinates and first score of the identified second detection box, and can determine the false detection filtering template for this part of the monitoring image based on the detection boxes of these falsely detected objects. This allows the detection boxes of falsely detected objects to be removed based on the false detection filtering template when performing target behavior detection on this part of the monitoring image, thus ensuring the accuracy of target behavior detection.

[0026] In the above technical solution, the step of determining the detection frame of the false detection object in the second detection frame according to the first score and the first coordinate specifically includes: determining the size and position of the second detection frame according to the first coordinate; and identifying the second detection frames of the same size and position in multiple historical monitoring images, and whose first score is less than the preset score, as the detection frames of the false detection object.

[0027] In this technical solution, the detection device can determine the detection frame of the false detection object in the second detection frame by comprehensively considering the position and size relationship of the second detection frame in multiple historical monitoring images, as well as the first score of the second detection frame. This ensures the accuracy of the determined detection frame of the false detection object, and consequently ensures the accuracy of the false detection filtering template set based on the detection frame of the false detection object in subsequent steps.

[0028] In the above technical solution, the step of processing the current monitoring image through a false detection filtering template to identify and remove false detection objects in the current monitoring image, and determining the first detection box of the target object in the current monitoring image, specifically includes: identifying the third detection box in the current monitoring image; determining the detection box of the false detection object in the third detection box according to the false detection filtering template; removing the detection box of the false detection object in the third detection box, and determining the remaining third detection box as the first detection box; wherein, the third detection box is the detection box of the false detection object and the target object in the current monitoring image.

[0029] In this technical solution, during the process of determining the first detection box, i.e., the detection box for the target object, the detection device removes detection boxes of falsely detected objects that resemble the target object through a false detection filtering template. This ensures the accuracy of the determined first detection box, and consequently, the accuracy of target behavior detection based on the first detection box.

[0030] In the above technical solution, the step of determining the detection box of the false detection object in the third detection box according to the false detection filtering template specifically includes: confirming whether a false detection filtering area is set in the current monitoring image; if it is confirmed that a false detection filtering area is set, using a template matching algorithm according to the false detection filtering template to identify the detection box of the false detection object in the false detection filtering area; or if it is confirmed that no false detection filtering area is set, using a template matching algorithm according to the false detection filtering template to identify the detection box of the false detection object in the current monitoring image.

[0031] In this technical solution, before determining the detection frame of a false detection object in the third detection frame, the detection device also needs to determine whether the user has set a false detection filtering area. If a false detection filtering area is set, only the detection frames of false detection objects within the false detection filtering area are filtered. This improves the efficiency of determining the detection frames of false detection objects and the efficiency of target behavior detection. If no false detection filtering area is set, the detection frames of false detection objects in the current monitoring image are filtered. This improves the accuracy of determining the detection frames of false detection objects and the accuracy of target behavior detection.

[0032] In the above technical solution, before determining the existence of target behavior, the detection method further includes: performing an overlap judgment on the first detection box and the target area to determine a first overlap rate; and determining that the first detection box is in the target area if the first overlap rate is greater than a preset overlap rate.

[0033] In this technical solution, the detection device can determine whether the first detection frame is in the target area by judging the relationship between the first overlap rate and the preset overlap rate, thus ensuring the accuracy of target behavior detection.

[0034] In the above technical solution, after determining that the target behavior exists, the detection method further includes: determining the location information of the target behavior, and sending the location information and the alarm information of the target behavior to the user terminal.

[0035] In this technical solution, after determining the existence of target behavior, the detection device can also send alarm information and location information of the target behavior to the user terminal. In this way, the user can promptly stop or handle the continued occurrence of the target behavior, effectively avoiding safety accidents caused by the target behavior.

[0036] In the above technical solution, before setting the false detection filtering template based on multiple historical monitoring images, the detection method further includes: obtaining user demand information and / or behavior detection scenario; and determining the target behavior to be detected based on the user demand information and / or behavior detection scenario.

[0037] In this technical solution, the detection device can automatically determine the target behavior to be identified based on the acquired user demand information and / or behavior detection scenario, thereby improving the detection efficiency of the target behavior.

[0038] According to a second aspect of the present invention, a target behavior detection device is provided, comprising: a first processing module for setting a false detection filtering template based on multiple historical monitoring images; a second processing module for identifying and removing falsely detected objects in the current monitoring image through the false detection filtering template, thereby determining a first detection box of a target object in the current monitoring image; and a third processing module for determining the existence of a target behavior when the first detection box is located in a target area in the current monitoring image.

[0039] In this technical solution, the aforementioned false detection objects refer to objects in the monitoring image that are easily identified as target objects but are not target objects, such as scarecrows or robots in a park; the aforementioned historical monitoring images refer to historical images captured from the monitoring video of the monitoring area; the aforementioned current monitoring image refers to the current image captured from the monitoring video of the monitoring area; the aforementioned false detection filtering template refers to a filtering template for false detection objects that are continuously falsely detected as target objects in the background of the monitoring image; the aforementioned first detection box refers to the frame line of the target object identified in the current monitoring image; the aforementioned target area refers to the violation area in the current monitoring image, such as a non-zebra crossing area; the aforementioned target behavior refers to the violation behavior, such as pedestrian violation behavior.

[0040] Specifically, the first processing module first determines a false detection filtering template for this portion of the surveillance images based on multiple historical surveillance images. Specifically, based on the multiple historical surveillance images, false detection objects that are easily identifiable as the target objects can be identified in this portion of the surveillance images. When identifying target objects in new surveillance images, the new surveillance images can be processed based on the aforementioned false detection objects. Therefore, the first processing module is able to determine the false detection filtering template based on multiple historical surveillance images.

[0041] Furthermore, the second processing module identifies the first detection box of the target object in the current monitoring image. Specifically, the second processing module first identifies the target object in the current monitoring image, then removes the false detection objects that were falsely detected as the target object from the identified target objects according to the aforementioned false detection filtering template, and finally determines the detection box of the remaining target object, namely the aforementioned first detection box.

[0042] Furthermore, the third processing module determines whether the first detection box is located within the target area set in the current monitoring image. Specifically, if it is determined that the first detection box is located within the target area, the third processing module determines that there is a target behavior in the monitoring image.

[0043] Specifically, if the first detection box is located in the target area, it indicates that the target object is in a violation area. Therefore, the third processing module can determine that there is target behavior in the monitoring image.

[0044] In this technical solution, the first processing module determines a false detection filtering template for a portion of the monitoring images based on multiple historical monitoring images. The second processing module removes false detection objects that are easily misidentified as target objects during the identification of target objects in the current monitoring image using the false detection filtering template, and determines a first detection box for all target objects. The third processing module determines whether target behavior exists based on the judgment result of whether the first detection box is within the target area. In the technical solution of this application, the second processing module can filter out false detection objects that are misidentified as target objects through the set false detection filtering template, solving the problem in related technologies where false detection of target objects occurs due to complex backgrounds in monitoring images, leading to false alarms of violations, and ensuring the accuracy of target behavior detection.

[0045] According to the third aspect of this application, another target behavior detection device is proposed, comprising: a memory storing a program or instructions; and a processor executing the program or instructions stored in the memory to implement the steps of the target behavior detection method proposed in the above-described technical solution of this application. Therefore, it has all the beneficial technical effects of the target behavior detection method proposed in the above-described technical solution of this application, which will not be elaborated further here.

[0046] According to the fourth aspect of this application, a readable storage medium is proposed, on which a program or instructions are stored. When the program or instructions are executed by a processor, they implement the steps of the target behavior detection method proposed in the above-described technical solution of this application. Therefore, it has all the beneficial technical effects of the target behavior detection method proposed in the above-described technical solution of this application, which will not be elaborated further here.

[0047] According to the fifth aspect of this application, an electronic device is proposed, including a target behavior detection device as proposed in the above-described technical solution of this invention, and / or a readable storage medium as proposed in the above-described technical solution of this invention. Therefore, the electronic device has all the beneficial effects of the target behavior detection device and / or the readable storage medium proposed in the above-described technical solution of this invention, which will not be repeated here.

[0048] According to the sixth aspect of this application, a computer program product is proposed, comprising a computer program that, when executed by a processor, implements the steps of the target behavior detection method proposed in the above-described technical solution of this application, and thus possesses all the beneficial technical effects of the target behavior detection method proposed in the above-described technical solution of this application, which will not be elaborated further here.

[0049] Additional aspects and advantages of this application will become apparent in the following description or may be learned by practice of this application. Attached Figure Description

[0050] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0051] Figure 1 One of the flowcharts of a target behavior detection method according to an embodiment of this application is shown;

[0052] Figure 2 A second schematic flowchart of a target behavior detection method according to an embodiment of this application is shown;

[0053] Figure 3 The third schematic flowchart of the target behavior detection method according to an embodiment of this application is shown;

[0054] Figure 4 The fourth schematic flowchart of the target behavior detection method according to an embodiment of this application is shown;

[0055] Figure 5 A schematic diagram of the current monitoring image in an embodiment of this application is shown;

[0056] Figure 6 The fifth flowchart illustrates the target behavior detection method according to an embodiment of this application;

[0057] Figure 7 This is a sixth schematic flowchart illustrating the target behavior detection method according to an embodiment of this application;

[0058] Figure 8 One of the schematic block diagrams of a target behavior detection device according to an embodiment of this application is shown;

[0059] Figure 9 This is a second schematic block diagram of a target behavior detection device according to an embodiment of this application;

[0060] Figure 10 A schematic block diagram of a target behavior detection system according to an embodiment of this application is shown. Detailed Implementation

[0061] To better understand the above-mentioned objectives, features, and advantages of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0062] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0063] The following is combined Figures 1 to 10The present application provides a detailed description of a target behavior detection method, apparatus, storage medium, and electronic device through specific embodiments and application scenarios.

[0064] Example 1:

[0065] Figure 1 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0066] S102, set false detection filtering templates based on multiple historical surveillance images;

[0067] S104, through the false detection filtering template, identify and remove false detection objects in the current monitoring image to determine the first detection box of the target object in the current monitoring image;

[0068] S106, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior.

[0069] It should be noted that the execution subject of the target behavior detection method proposed in this application can be a target behavior detection device. In order to more clearly explain the target behavior detection method proposed in this application, the following embodiments use the target behavior detection device as the execution subject of the target behavior detection method for illustrative purposes.

[0070] In this embodiment, the aforementioned false detection objects refer to objects in the monitoring image that are easily identified as target objects but are not target objects, such as scarecrows or robots in a park; the aforementioned historical monitoring images refer to historical images captured from the monitoring video of the monitoring area; the aforementioned current monitoring image refers to the current image captured from the monitoring video of the monitoring area; the aforementioned false detection filtering template refers to a filtering template for false detection objects that are continuously falsely detected as target objects in the background of the monitoring image; the aforementioned first detection box refers to the frame line of the target object identified in the current monitoring image; the aforementioned target area refers to the violation area in the current monitoring image, such as a non-zebra crossing area; the aforementioned target behavior refers to a violation, such as a pedestrian violation.

[0071] Specifically, the detection device first determines a false detection filtering template for that portion of the surveillance images based on multiple historical surveillance images. Specifically, based on these historical images, it can identify falsely detected objects in that portion of the surveillance images that are easily mistaken for the target object. When identifying the target object in a new portion of the surveillance images, the new images can be processed based on these falsely detected objects. For example, when detecting pedestrians, it can identify objects in that portion of the surveillance images that are easily mistaken for people, such as scarecrows or robots, based on multiple historical images. Then, when a new surveillance image is obtained, the scarecrows and robots in the new image can be removed. Therefore, the detection device can determine a false detection filtering template based on multiple historical surveillance images.

[0072] Furthermore, the detection device identifies the first detection box of the target object in the current monitoring image. Specifically, the detection device first identifies the target object in the current monitoring image, then removes falsely detected objects that were mistakenly identified as the target object from the identified target objects according to the aforementioned false detection filtering template, and finally determines the detection boxes of the remaining target objects, namely the aforementioned first detection boxes.

[0073] Furthermore, the detection device determines whether the first detection frame is located within the target area set in the current monitoring image. Specifically, if it is determined that the first detection frame is located within the target area, the detection device determines that there is a target behavior in the monitoring image.

[0074] Specifically, if the first detection box is located in the target area, it indicates that the target object is in a violation area. Therefore, the detection device can determine that there is target behavior in the monitoring image.

[0075] For example, the process of this application is explained by taking the false detection object as an object resembling a pedestrian, the target object as a pedestrian, the target area as a non-zebra crossing area, and the target behavior as a pedestrian violation.

[0076] Specifically, the detection device first identifies objects that are easily misidentified as pedestrians based on historical surveillance images near the zebra crossing, and then sets up false detection filtering templates based on these objects. In other words, the detection device uses image recognition technology to identify pedestrians or similar objects in historical surveillance images and can determine the probability score of these objects as pedestrians. By identifying multiple historical surveillance images, the detection device can identify objects with low probability scores that do not change across multiple historical surveillance images as objects that are mistakenly identified as pedestrians.

[0077] Furthermore, the detection device identifies pedestrians in the current surveillance image near the zebra crossing and removes non-pedestrian objects from the identified pedestrians according to the false detection filtering template, so as to determine the first detection box for all pedestrians in the current surveillance image.

[0078] Furthermore, the detection device determines whether the pedestrian's first detection frame is located in a non-zebra crossing area in the current monitoring image. If it is located in a non-zebra crossing area, it is determined that there is a pedestrian violation in the current monitoring image.

[0079] For example, the target behavior detection method proposed in this application can be used in intelligent security scenarios in parks or factories. Specifically, the target behaviors included in parks or factories mainly include production line crossing behavior, park intrusion behavior, pedestrian violation behavior, and vehicle illegal parking behavior.

[0080] In this embodiment, the detection device can determine a false detection filtering template for a portion of the monitoring images based on multiple historical monitoring images. This template removes falsely identified objects that are easily mistaken for target objects during the identification process in the current monitoring image. A first detection box is then determined for all target objects. Based on the judgment result of whether the first detection box is within the target area, the existence of target behavior can be determined. In this embodiment, the detection device can filter out falsely identified objects using the set false detection filtering template, solving the problem in related technologies where complex backgrounds in monitoring images lead to false detection of target objects and false alarms of violations, thus ensuring the accuracy of target behavior detection.

[0081] Figure 2 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0082] S202, Identify the second detection box in multiple historical surveillance images, and determine the first score and first coordinates of the second detection box;

[0083] S204, Determine the detection frame of the falsely detected object in the second detection frame based on the first score and the first coordinate;

[0084] S206, Set the false detection filtering template according to the detection frame of the false detection object;

[0085] S208, through the false detection filtering template, identify and remove false detection objects in the current monitoring image to determine the first detection box of the target object in the current monitoring image;

[0086] S210, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior.

[0087] The second detection box is the detection box of the false detection object and the target object in multiple historical monitoring images, and the first score is the probability score of the second detection box being the first detection box.

[0088] In this embodiment, the second detection box represents all detection boxes of the target object and falsely detected objects similar to the target object in the multiple historical surveillance images identified; the first score represents the score of the detection box that is the target object; and the first coordinate represents the position and size information of the detection box in the historical surveillance image.

[0089] Specifically, the process of determining the false detection filtering template is as follows: the detection device identifies and determines all the second detection boxes included in multiple of the above-mentioned historical monitoring images, and determines the first coordinates and first scores of these second detection boxes.

[0090] Specifically, the detection device can determine the second detection box, as well as the first coordinates and the first score of the second detection box, by inputting multiple of the aforementioned historical surveillance images into a pre-trained neural network model or a YOLOv5 (object detection) network model.

[0091] Furthermore, the detection device uses the first coordinates and the first score to solve for the detection frame of the false detection object in the second detection frame. Specifically, the detection device can determine whether the position and size of the second detection frame have changed in multiple historical surveillance images based on the first coordinates, and can determine the probability that the second detection frame is a false detection object based on the first score. Based on these conditions, it can further determine whether the second detection frame is a false detection object.

[0092] Furthermore, the detection device sets a false detection filtering template for the current monitoring image based on the detection frames of these false detection objects.

[0093] In this embodiment, the detection device can determine the detection boxes of falsely detected objects included in the second detection box based on the first coordinates and first score of the identified second detection box, and can determine the false detection filtering template for this part of the monitoring image based on the detection boxes of these falsely detected objects. This allows the detection boxes of falsely detected objects to be removed based on the false detection filtering template when performing target behavior detection on this part of the monitoring image, thus ensuring the accuracy of target behavior detection.

[0094] In the above embodiments, the step of determining the detection box of the false detection object in the second detection box according to the first score and the first coordinate specifically includes: determining the size and position of the second detection box according to the first coordinate; and determining the second detection box of the same size and position in multiple historical monitoring images, and whose first score is less than the preset score, as the detection box of the false detection object.

[0095] In this embodiment, the process of determining the detection frame of the false detection object in the second detection frame, that is, determining the false detection object frame, is as follows: the detection device first calculates the position and size of the second detection frame in the monitoring image based on the first coordinates.

[0096] Furthermore, the detection device determines whether the position and size of the second detection frame are the same in multiple historical monitoring images, and the relationship between the first score and the preset score of the second detection frame with the same position and size.

[0097] Specifically, when the detection device determines that the position and size of the second detection frame are the same in multiple historical monitoring images and the first score is less than the preset score, the detection frame is identified as a false detection object.

[0098] Specifically, if the position and size of the second detection box are the same in multiple of the aforementioned historical surveillance images, and the first score of the second detection box is less than the aforementioned preset score, it indicates that the second detection box is in a stationary state, and the second detection box is likely not the detection box of the target object. Therefore, the second detection box can be identified as the detection box of the false detection object.

[0099] For example, the above preset score is set in advance according to user needs, usually set to 0.5.

[0100] In this embodiment, the detection device can determine the detection frame of a false positive in the second detection by comprehensively considering the position and size relationship of the second detection frame in multiple historical surveillance images, as well as the first score of the second detection frame. This ensures the accuracy of the determined detection frames of false positives, thereby guaranteeing the accuracy of the false positive filtering template set based on the detection frames of the false positives in subsequent steps.

[0101] Figure 3 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0102] S302, set false detection filtering templates based on multiple historical surveillance images;

[0103] S304, Identify the third detection box in the current monitoring image;

[0104] S306, Determine the detection frame of the false detection object in the third detection frame according to the false detection filter template;

[0105] S308, Remove the detection boxes of falsely detected objects in the third detection box, and determine the remaining third detection boxes as the first detection boxes;

[0106] S310, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior.

[0107] The third detection box is the detection box for falsely detected objects and target objects in the current monitoring image.

[0108] In this embodiment, the second detection box represents all detection boxes of the target object and falsely detected objects similar to the target object in the current monitoring image.

[0109] Specifically, the process of determining the first detection box is as follows: the detection device first identifies all the third detection boxes included in the current monitoring image. Specifically, the detection device inputs the current monitoring image into a pre-trained neural network model to determine the third detection boxes.

[0110] Furthermore, the detection device uses the aforementioned false detection filtering template to determine the detection frames of falsely detected objects included in the third detection frame. Specifically, the false detection filtering template can clearly identify the location of the detection frames of falsely detected objects in the background of the current monitoring image, and based on this location, it can filter out which detection frames in the third detection frame are falsely detected objects.

[0111] Furthermore, the detection device deletes the detection frames containing falsely detected objects included in the third detection frame and determines the remaining third detection frame as the first detection frame representing the target object.

[0112] In this embodiment, during the process of determining the first detection box, i.e., the detection box for the target object, the detection device removes detection boxes of falsely detected objects that resemble the target object using a false detection filtering template. This ensures the accuracy of the determined first detection box, and consequently, the accuracy of target behavior detection based on the first detection box.

[0113] Figure 4 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0114] S402, set false detection filtering templates based on multiple historical surveillance images;

[0115] S404, identify the third detection box in the current monitoring image;

[0116] S406, Confirm whether there is a false detection filtering area set in the current monitoring image;

[0117] S408, if it is confirmed that a false detection filtering area is set, the template matching algorithm is used to identify the detection box of the false detection object in the false detection filtering area according to the false detection filtering template.

[0118] S410, if it is confirmed that no false detection filtering area is set, the template matching algorithm is used according to the false detection filtering template to identify the detection box of the false detection object in the current monitoring image;

[0119] S412, Remove the detection boxes of falsely detected objects in the third detection box, and determine the remaining third detection box as the first detection box;

[0120] S414, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior.

[0121] In this embodiment, the process of determining the detection frame that includes false detection objects in the third detection frame is as follows: the detection device first determines whether the user has set a false detection filtering area in the current monitoring screen.

[0122] Furthermore, when the detection device determines that the user has set a false detection filtering area, the detection device uses the false detection filtering template to identify only the detection frames of false detection objects in the third detection frame of the aforementioned false detection filtering area.

[0123] Specifically, if the user sets the above-mentioned false detection filtering area, it indicates that the false detection filtering area is a high-incidence area for false detections. Therefore, the detection device only needs to filter the detection boxes of false detection objects in the third detection box of the false detection filtering area. This is beneficial to improving the determination efficiency of the detection boxes of false detection objects and improving the efficiency of target behavior detection.

[0124] Furthermore, when the detection device determines that the user has not set a false detection filtering area, the detection device identifies the detection frames of all false detection objects in the third detection frames in the current monitoring image through the false detection filtering template.

[0125] Specifically, if the user does not set the aforementioned false detection filtering area, it means that all areas in the current monitoring image may be falsely detected. Therefore, the detection device needs to filter the detection boxes of all falsely detected objects in the third detection boxes in the current monitoring image. This helps to improve the accuracy of determining the detection boxes of falsely detected objects and improves the accuracy of target behavior detection.

[0126] For example, taking a pedestrian as the target object and a pedestrian violation as the target behavior, the schematic diagram of the current monitoring image is as follows: Figure 5 As shown, Figure 5 In the diagram, box A represents the target area, box B represents the false detection filtering area, box C represents the identified pedestrian detection box (i.e., the third detection box mentioned above), and box D represents the detection box of the false detection object in the false detection filtering template.

[0127] Specifically, when filtering the detection boxes of falsely detected objects in the current monitoring image according to the false detection filtering template, it is necessary to first determine whether there is a frame B representing the false detection filtering area in the current monitoring image. If it is determined that there is a frame B, only the detection boxes of falsely detected objects in the third detection boxes within the area where frame B is located need to be filtered. If it is determined that there is no frame B, it is necessary to filter the detection boxes of falsely detected objects in all the third detection boxes in the current monitoring image. It should be noted that this application uses a template matching algorithm to identify the detection boxes of falsely detected objects in the third detection boxes. When identifying the detection boxes of falsely detected objects using the template matching algorithm, if there is a pixel deviation in the position coordinates of the detection boxes, it will lead to errors in the results obtained by the template matching algorithm. Therefore, in the embodiments of this application, the pixel size of the false detection filtering template is limited. Taking the pixel length and width of the false detection filtering template as 60 and 20 as an example, the pixel length and width of the current monitoring image need to be set to 72 and 24, that is, the size offset between the pixel size of the filtering template and the pixel size of the current monitoring image is set to 20%. This size offset can be adjusted according to the actual situation, but it needs to be kept between 10% and 100%.

[0128] In this embodiment, before determining the detection frame of a false detection object in the third detection frame, the detection device also needs to determine whether the user has set a false detection filtering area. If a false detection filtering area is set, only the detection frames of false detection objects within the false detection filtering area are filtered. This improves the efficiency of determining the detection frames of false detection objects and the efficiency of target behavior detection. If no false detection filtering area is set, the detection frames of false detection objects in the current monitoring image are filtered. This improves the accuracy of determining the detection frames of false detection objects and the accuracy of target behavior detection.

[0129] Figure 6 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0130] S602, sets false detection filtering templates based on multiple historical surveillance images;

[0131] S604, through the false detection filtering template, identifies and removes false detection objects in the current monitoring image to determine the first detection box of the target object in the current monitoring image;

[0132] S606, perform overlap judgment on the first detection box and the target area, and determine the first overlap rate;

[0133] S608, if the first overlap rate is greater than the preset overlap rate, determine that the first detection box is in the target area;

[0134] S610, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior.

[0135] In this embodiment, the detection device specifically determines whether a target behavior exists by judging whether the first detection box is located in the target area of ​​the current monitoring image. Specifically, the method for judging whether the first detection box is located in the target area is as follows: the detection device determines the first overlap rate between the first detection box and the target area by performing an overlap judgment on the first detection box and the target area.

[0136] For example, the above-mentioned method of determining overlap can be any one of the following: midpoint overlap of the baseline, complete overlap, and partial overlap.

[0137] Furthermore, the detection device determines the relationship between the first overlap rate and the preset overlap rate. Specifically, when it is determined that the first overlap rate is greater than the preset overlap rate, the first detection frame corresponding to the first overlap rate is located in the target area, thus confirming the existence of the target behavior.

[0138] In this embodiment, the detection device can determine whether the first detection box is in the target area by judging the relationship between the first overlap rate and the preset overlap rate, thus ensuring the accuracy of target behavior detection.

[0139] Figure 7 A flowchart illustrating a method for detecting target behavior according to an embodiment of this application is shown. The detection method includes:

[0140] S702 sets false detection filtering templates based on multiple historical surveillance images;

[0141] S704 identifies and removes false detection objects in the current monitoring image through a false detection filtering template, thereby determining the first detection box of the target object in the current monitoring image;

[0142] S706, if the first detection box is located in the target area of ​​the current monitoring image, it is determined that there is a target behavior;

[0143] S708 determines the location information of the target behavior and sends the location information and alarm information indicating the presence of the target behavior to the user terminal.

[0144] In this embodiment, after determining the current monitoring image, the detection device also needs to determine the location information of the target behavior. Specifically, the detection device can calculate the location information based on the first coordinates of the first detection box in the target area.

[0145] Furthermore, the detection device sends the aforementioned location information and alarm information indicating that the target behavior has occurred to the user terminal.

[0146] In this embodiment, after determining that a target behavior exists, the detection device can also send alarm information and location information of the target behavior to the user terminal. In this way, the user can promptly stop or handle the continued occurrence of the target behavior, effectively avoiding safety accidents caused by the target behavior.

[0147] In the above embodiments, before setting the false detection filtering template based on multiple historical monitoring images, the detection method further includes: obtaining user demand information and / or behavior detection scenarios; and determining the target behavior to be detected based on the user demand information and / or behavior detection scenarios.

[0148] In this embodiment, the aforementioned user requirement information represents the information of the target behavior that the user needs to detect, and the aforementioned behavior recognition scenario represents the scenario of the target behavior to be identified, such as a park or workshop.

[0149] Specifically, before setting the false detection filtering template based on multiple historical surveillance images, the detection device also needs to determine what the target behavior to be identified is based on the aforementioned user requirement information and / or the aforementioned behavior recognition scenario, thereby determining what the target object to be identified is and the location of the target area to be identified.

[0150] In this embodiment, the detection device can automatically determine the target behavior to be identified based on the acquired user demand information and / or behavior detection scenario, thereby improving the detection efficiency of the target behavior.

[0151] Example 2:

[0152] Figure 8 A schematic block diagram of a target behavior detection device according to an embodiment of this application is shown. The target behavior detection device 800 includes: a first processing module 802, used to set a false detection filtering template based on multiple historical monitoring images; a second processing module 804, used to identify and remove false detection objects in the current monitoring image through the false detection filtering template, so as to determine a first detection box of the target object in the current monitoring image; and a third processing module 806, used to determine that a target behavior exists when the first detection box is located in the target area of ​​the current monitoring image.

[0153] In this embodiment, the aforementioned false detection objects refer to objects in the monitoring image that are easily identified as target objects but are not target objects, such as scarecrows or robots in a park; the aforementioned historical monitoring images refer to historical images captured from the monitoring video of the monitoring area; the aforementioned current monitoring image refers to the current image captured from the monitoring video of the monitoring area; the aforementioned false detection filtering template refers to a filtering template for false detection objects that are continuously falsely detected as target objects in the background of the monitoring image; the aforementioned first detection box refers to the frame line of the target object identified in the current monitoring image; the aforementioned target area refers to the violation area in the current monitoring image, such as a non-zebra crossing area; the aforementioned target behavior refers to a violation, such as a pedestrian violation.

[0154] Specifically, the first processing module 802 first determines a false detection filtering template for this portion of the monitoring images based on multiple historical monitoring images. Specifically, based on the multiple historical monitoring images, false detection objects that are easily identified as the target objects in this portion of the monitoring images can be determined. When identifying target objects in new monitoring images, the new monitoring images can be processed based on the aforementioned false detection objects. Therefore, the first processing module 802 is able to determine the false detection filtering template based on multiple historical monitoring images.

[0155] Furthermore, the second processing module 804 identifies the first detection box of the target object in the current monitoring image. Specifically, the second processing module 804 first identifies the target object in the current monitoring image, then removes the false detection objects that were falsely detected as the target object from the identified target objects according to the aforementioned false detection filtering template, and finally determines the detection box of the remaining target object, namely the aforementioned first detection box.

[0156] Furthermore, the third processing module 806 determines whether the first detection box is located within the target area set in the current monitoring image. Specifically, if it is determined that the first detection box is located within the target area, the third processing module 806 determines that there is a target behavior in the monitoring image.

[0157] Specifically, if the first detection box is located in the target area, it indicates that the target object is in a violation area. Therefore, the third processing module 806 can determine that there is target behavior in the monitoring image.

[0158] In this embodiment, the first processing module 802 can determine a false detection filtering template for a portion of the monitoring images based on multiple historical monitoring images. The second processing module 804 can remove false detection objects that are easily misidentified as target objects during the identification of target objects in the current monitoring image using the false detection filtering template, and determine a first detection box for all target objects. The third processing module 806 can determine whether there is a target behavior based on the judgment result of whether the first detection box is within the target area. In this embodiment of the application, the second processing module 804 can filter out false detection objects that are misidentified as target objects through the set false detection filtering template, which solves the problem in related technologies where the complex background of the monitoring image leads to false detection of target objects and false alarms of violations, thus ensuring the accuracy of target behavior detection.

[0159] In the above embodiments, the first processing module 802 is specifically used to identify a second detection box in multiple historical monitoring images and determine a first score and a first coordinate of the second detection box; determine the detection box of a false detection object in the second detection box based on the first score and the first coordinate; and set a false detection filtering template based on the detection box of the false detection object; wherein, the second detection box is the detection box of the false detection object and the target object in multiple historical monitoring images, and the first score is the probability score of the second detection box being the first detection box.

[0160] In this embodiment, the first processing module 802 can determine the detection boxes of falsely detected objects included in the second detection box based on the first coordinates and first score of the identified second detection box, and can determine the false detection filtering template for this part of the monitoring image based on the detection boxes of these falsely detected objects, so that when performing target behavior detection on this part of the monitoring image, the detection boxes of falsely detected objects can be removed according to the false detection filtering template, thus ensuring the accuracy of target behavior detection.

[0161] In the above embodiment, the first processing module 802 is specifically used to determine the size and position of the second detection box according to the first coordinate; and to determine the second detection box with the same size and position in multiple historical monitoring images and whose first score is less than the preset score as the detection box of the false detection object.

[0162] In this embodiment, the first processing module 802 can determine the detection box of the false detection object in the second detection by comprehensively considering the position and size relationship of the second detection box in multiple historical monitoring images and the first score of the second detection box. This ensures the accuracy of the determined detection box of the false detection object, thereby ensuring the accuracy of the false detection filtering template set based on the detection box of the false detection object in subsequent steps.

[0163] In the above embodiment, the second processing module 804 is specifically used to identify the third detection box in the current monitoring image; determine the detection box of the false detection object in the third detection box according to the false detection filtering template; remove the detection box of the false detection object in the third detection box, and determine the remaining third detection box as the first detection box; wherein, the third detection box is the detection box of the false detection object and the target object in the current monitoring image.

[0164] In this embodiment, during the process of determining the first detection box, i.e., the detection box for the target object, the second processing module 804 removes detection boxes of falsely detected objects that resemble the target object using a false detection filtering template. This ensures the accuracy of the determined first detection box, and consequently, the accuracy of target behavior detection based on the first detection box.

[0165] In the above embodiments, the second processing module 804 is specifically used to confirm whether a false detection filtering area is set in the current monitoring image; if it is confirmed that a false detection filtering area is set, a template matching algorithm is used according to the false detection filtering template to identify the detection box of the false detection object in the false detection filtering area; or if it is confirmed that no false detection filtering area is set, a template matching algorithm is used according to the false detection filtering template to identify the detection box of the false detection object in the current monitoring image.

[0166] In this embodiment, before determining the detection boxes of false detection objects in the third detection box, the second processing module 804 also needs to determine whether the user has set a false detection filtering area. If a false detection filtering area is set, only the detection boxes of false detection objects in the false detection filtering area are filtered, thus improving the efficiency of determining the detection boxes of false detection objects and improving the efficiency of target behavior detection. If no false detection filtering area is set, the detection boxes of false detection objects in the current monitoring image are filtered, thus improving the accuracy of determining the detection boxes of false detection objects and improving the accuracy of target behavior detection.

[0167] In the above embodiments, the third processing module 806 is further configured to perform an overlap judgment on the first detection box and the target area, and determine a first overlap rate; if the first overlap rate is greater than a preset overlap rate, determine that the first detection box is in the target area.

[0168] In this embodiment, the third processing module 806 can determine whether the first detection box is in the target area by judging the relationship between the first overlap rate and the preset overlap rate, thus ensuring the accuracy of target behavior detection.

[0169] In the above embodiments, the third processing module 806 is further configured to determine the location information of the target behavior and send the location information and alarm information indicating the existence of the target behavior to the user terminal.

[0170] In this embodiment, after determining that a target behavior exists, the third processing module 806 can also send alarm information and location information of the target behavior to the user terminal. In this way, the user can stop or handle the target behavior from continuing to occur in a timely manner, effectively avoiding safety accidents caused by the target behavior.

[0171] In the above technical solution, the first processing module is also used to acquire user demand information and / or behavior detection scenario; the first processing module is also used to determine the target behavior to be detected based on the user demand information and / or behavior detection scenario.

[0172] In this technical solution, the first processing module can automatically determine the target behavior to be identified based on the acquired user demand information and / or behavior detection scenario, thereby improving the detection efficiency of the target behavior.

[0173] Example 3:

[0174] Figure 9 A schematic block diagram of another target behavior detection device 900 according to an embodiment of this application is shown. The target behavior detection device 900 includes: a memory 902, in which a program or instructions are stored; and a processor 904, which executes the program or instructions stored in the memory 902 to implement the steps of the target behavior detection method proposed in the above embodiments of this application. Therefore, it has all the beneficial technical effects of the target behavior detection method proposed in the above embodiments of this application, and will not be described in detail here.

[0175] Example 4:

[0176] According to the fourth embodiment of this application, a readable storage medium is proposed, on which a program or instructions are stored. When the program or instructions are executed by a processor, they implement the steps of the target behavior detection method proposed in the above embodiments of this application, and thus have all the beneficial technical effects of the target behavior detection method proposed in the above embodiments of this application, which will not be elaborated further here.

[0177] Example 5:

[0178] According to the fifth embodiment of this application, an electronic device is proposed, including a target behavior detection device as proposed in the above embodiments of the present invention, and / or a readable storage medium as proposed in the above embodiments of the present invention. Therefore, the electronic device has all the beneficial effects of the target behavior detection device and / or the readable storage medium proposed in the above embodiments of the present invention, which will not be repeated here.

[0179] Example 6:

[0180] According to the sixth embodiment of this application, a computer program product is proposed, including a computer program. When the computer program is executed by a processor, it implements the steps of the target behavior detection method proposed in the above embodiments of this application, and thus has all the beneficial technical effects of the target behavior detection method proposed in the above embodiments of this application, which will not be elaborated further here.

[0181] Example 7:

[0182] For example, the target behavior detection method proposed in the above embodiments of this application can also be achieved through methods such as... Figure 10 The target behavior detection system shown in this embodiment is implemented by combining... Figure 10 The target behavior detection system shown is illustrated using a factory park scenario to illustrate the target behavior detection method.

[0183] Specifically, the data module can collect current monitoring images of the factory park through surveillance cameras 1 to n, and can send the collected images to the target behavior detection system through the communication module.

[0184] according to Figure 10 As can be seen, users can configure the target behavior to be detected and the target object to be detected through the detection behavior configuration module of the detection system. For example: if the target behavior is set to pedestrian violation, the target object to be detected is pedestrian; if the target behavior is set to vehicle driving in the wrong direction, the target object to be detected is vehicle; if the target behavior is set to stacked items, the target object to be detected is items; if the target behavior is set to cross the production line, the target object to be detected is pedestrian, etc.

[0185] according to Figure 10 As can be seen, users can configure the monitoring area of ​​the camera in the data acquisition module through the detection behavior configuration module of the detection system, or select the corresponding camera.

[0186] Specifically, the detection system stores a false detection filtering template for a certain target behavior, or the detection system can determine the false detection filtering template based on the acquired monitoring image, and can filter false detections in the acquired current monitoring image based on the false detection filtering template corresponding to the target behavior to be detected, so as to continuously detect the target behavior.

[0187] Furthermore, upon detecting the presence of a target behavior, the detection system can interact with the administrator terminal, i.e., it can report the presence of the target behavior to the administrator terminal, and the administrator terminal can report the processing result of the target behavior back to the detection system. In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance, unless otherwise expressly specified and limited; the terms "connection," "installation," and "fixing," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0188] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0189] Furthermore, the embodiments of this application can be combined with each other, but only if they are based on what those skilled in the art can do. If the combination of embodiments is contradictory or cannot be implemented, it should be considered that such combination of embodiments does not exist and is not within the scope of protection claimed by this application.

[0190] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of detecting a target behavior, characterized by, include: Set up false detection filtering templates based on multiple historical surveillance images; The false detection filtering template is used to identify and remove falsely detected objects in the current monitoring image to determine the first detection box of the target object in the current monitoring image. If the first detection box is located in the target area of ​​the current monitoring image, it is determined that the target behavior exists; Before determining the existence of the target behavior, the detection method further includes: The overlap between the first detection box and the target region is determined to establish a first overlap rate. If the first overlap rate is greater than the preset overlap rate, it is determined that the first detection box is in the target area; Among them, the method of overlapping judgment is any one of the following: overlapping of the midpoint of the baseline, complete overlapping, and partial overlapping judgment; The step of setting a false detection filtering template based on multiple historical surveillance images specifically includes: Multiple historical surveillance images are input into a pre-trained neural network model or object detection network model to identify a second detection box in the multiple historical surveillance images and determine the first score and first coordinates of the second detection box; The size and position of the second detection frame are determined based on the first coordinates; The second detection boxes in multiple historical surveillance images that are the same size and position and whose first score is less than a preset score are identified as the false detection objects; the false detection filtering template is set according to the false detection objects. Wherein, the second detection box is the detection box of the false detection object and the target object in the multiple historical monitoring images, and the first score is the probability score of the second detection box being the first detection box.

2. The method of claim 1, wherein The current monitoring image is processed using the false detection filtering template to identify and remove falsely detected objects in the current monitoring image, thereby determining the first detection box of the target object in the current monitoring image. Specifically, this includes: Identify the third detection box in the current monitoring image; The detection frame of the false detection object in the third detection frame is determined according to the false detection filtering template; Remove the detection boxes of the falsely detected objects in the third detection box, and determine the remaining third detection boxes as the first detection boxes; The third detection box is the detection box for the falsely detected object and the target object in the current monitoring image.

3. The method of claim 2, wherein The step of determining the detection frame of the false detection object in the third detection frame according to the false detection filtering template specifically includes: Confirm whether a false detection filtering area is set in the current monitoring image; If the false detection filtering area is confirmed to be set, a template matching algorithm is used to identify the detection box of the false detection object in the false detection filtering area according to the false detection filtering template; or If it is confirmed that no false detection filtering area is set, a template matching algorithm is used to identify the detection box of the false detection object in the current monitoring image based on the false detection filtering template.

4. The method of claim 1, wherein After determining that the target behavior exists, the detection method further includes: The location information of the target behavior is determined, and the location information and the alarm information of the target behavior are sent to the user terminal.

5. The method of claim 1, wherein Before setting the false detection filtering template based on multiple historical surveillance images, the detection method further includes: Acquiring user needs information and / or detecting behavior scenarios; Based on the user demand information and / or the behavior detection scenario, the target behavior to be detected is determined.

6. A device for detecting a target behavior, characterized by include: The first processing module is used to set false detection filtering templates based on multiple historical surveillance images; The second processing module is used to identify and remove false detection objects in the current monitoring image through the false detection filtering template, so as to determine the first detection box of the target object in the current monitoring image; The third processing module is used to determine the existence of the target behavior when the first detection box is located in the target area of ​​the current monitoring image; The third processing module is also used to determine the overlap between the first detection box and the target area and to determine the first overlap rate; if the first overlap rate is greater than the preset overlap rate, it is determined that the first detection box is in the target area; the overlap determination method is any one of the following: midpoint overlap of the baseline, complete overlap, and partial overlap determination. The first processing module is specifically used to input multiple historical surveillance images into a pre-trained neural network model or object detection network model to identify second detection boxes in the multiple historical surveillance images, and determine the first score and first coordinates of the second detection box; determine the detection boxes of false detection objects in the second detection box based on the first score and first coordinates; and set a false detection filtering template based on the detection boxes of false detection objects; wherein, the second detection box is the detection box of false detection objects and target objects in the multiple historical surveillance images, and the first score is the probability score of the second detection box being the first detection box; The first processing module is specifically used to determine the size and position of the second detection box based on the first coordinates; and to identify the second detection boxes in multiple historical monitoring images that are the same size and position and whose first score is less than the preset score as false detection objects.

7. A device for detecting a target behavior, characterized by include: A memory and a processor, the memory storing a program, the processor executing the program to implement the steps of the method for detecting the target behavior as described in any one of claims 1 to 5.

8. A readable storage medium, characterized by, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the detection method for the target behavior as described in any one of claims 1 to 5.

9. An electronic device, comprising: include: The target behavior detection device as described in claim 6 or 7; and / or The readable storage medium as described in claim 8.

10. A computer program product, characterised in that, It includes a computer program that, when executed by a processor, implements the steps of the method for detecting the target behavior as described in any one of claims 1 to 5.

Images