Human destruction behavior identification method and device, electronic equipment and storage medium

Abstract

The invention discloses a man-made destruction behavior identification method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining a to-be-identified power grid monitoring video; extracting a target image frame only containing a detection target according to the to-be-identified power grid monitoring video; for each target image frame, inputting the current target image frame into a preset posture recognition model, so that the preset posture recognition model generates key point information in the current target image frame according to the current target image frame; splicing all key point information to generate a key point sequence; and inputting the key point sequence into a behavior judgment model, so that the behavior judgment model judges whether a man-made destruction behavior exists in the key point sequence or not according to the key point sequence, and a behavior judgment result is generated. According to the invention, the accuracy of human destruction behavior identification can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method, device, electronic equipment and storage medium for identifying sabotage behavior. Background Art

[0002] In the security monitoring of power equipment and facilities, timely identification of external force destructive behavior is the key to ensuring the stability and safe operation of the power system. With the continuous expansion of the scale of power grid equipment and the increasing diversity of external threats, traditional manual monitoring and simple image recognition methods can no longer meet the needs of modern power grid security monitoring. Through the analysis of power grid monitoring videos, the on-site environment can be monitored and evaluated in real time, so as to timely detect potential destructive behaviors, prevent accidents, and improve the safety and reliability of the power grid. Especially in the security protection of key facilities such as substations and transmission lines, early and accurate identification and positioning of destructive behaviors can issue alarms before the threat causes serious damage, and take effective emergency response measures, thereby reducing damage to the power system and reducing maintenance costs.

[0003] However, most of the existing technologies for identifying vandalism rely on manual observation or basic image recognition, which makes it difficult to automatically identify vandalism and cannot extract abnormal behavior features from complex video data in a timely and accurate manner, resulting in insufficient accuracy in the identification of vandalism. Manual observation is highly dependent on manpower, and has problems such as strong subjectivity and fatigue affecting judgment. During long-term monitoring, manual observation is prone to distraction, missed judgments or misjudgments, making it difficult to detect sudden vandalism in a timely manner. Basic image recognition technologies, such as simple target detection or motion trajectory analysis, can only perform basic recording and static analysis of video streams, and cannot deeply understand the behavioral characteristics of people in the video. For example, the system may only recognize that someone appears in the picture, but cannot further determine whether their behavior is normal or abnormal. Summary of the invention

[0004] The embodiment of the present invention provides a method, device, electronic device and storage medium for identifying sabotage behavior. By implementing the present invention, automatic identification of sabotage behavior can be achieved, and the accuracy of identification of sabotage behavior can be improved.

[0005] An embodiment of the present invention provides a method for identifying sabotage behavior, including:

[0006] Obtain the power grid monitoring video to be identified;

[0007] Extracting target image frames containing only the detection target according to the power grid monitoring video to be identified;

[0008] For each target image frame, inputting the current target image frame into a preset gesture recognition model, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame;

[0009] All key point information is spliced ​​to generate a key point sequence;

[0010] The key point sequence is input into the behavior judgment model so that the behavior judgment model judges whether there is human sabotage behavior in the key point sequence according to the key point sequence, and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; each first training sample includes a training key point sequence generated by a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

[0011] Furthermore, the step of extracting a target image frame containing only the detection target according to the power grid monitoring video to be identified includes:

[0012] Perform frame sampling on the power grid monitoring video to be identified to generate a number of image frames;

[0013] For each image frame, the current image frame is input into the target detection model, so that the target detection model generates the position information of the detection target in the current image frame according to the current image frame; wherein the target detection model is trained by a plurality of second training samples; each second training sample includes an image frame sample and a picture label corresponding to the image frame sample; the picture label includes the position information of the detection target;

[0014] According to the position information of the detected target in each image frame, a target image frame corresponding to each image frame and containing only the detected target is generated.

[0015] Furthermore, the frame sampling of the power grid monitoring video to be identified to generate a plurality of image frames includes:

[0016] According to the preset duration, the power grid monitoring video to be identified is divided into several video clips;

[0017] According to a preset sampling frequency, each video clip is frame sampled to generate a number of original image frames;

[0018] Performing Gaussian denoising on each original image frame to generate a number of denoised image frames;

[0019] Image enhancement is performed on each denoised image frame to generate a plurality of image frames.

[0020] Furthermore, the training of the behavior judgment model includes:

[0021] Obtaining a number of first training samples;

[0022] Input each first training sample into the behavior judgment model in sequence, and train the behavior judgment model until a preset first training number is reached; wherein, the behavior judgment model outputs a predicted behavior judgment result corresponding to the first training sample each time it receives a first training sample; calculates a first loss function value based on the predicted behavior judgment result and the corresponding behavior judgment label; and updates the behavior judgment model based on the first loss function value.

[0023] Furthermore, the first training sample is obtained by:

[0024] Obtain power grid monitoring video samples and corresponding behavior judgment labels;

[0025] Perform frame sampling on power grid monitoring video samples to generate several training image frames;

[0026] For each training image frame, input the current training image frame into the target detection model, so that the target detection model generates the position information of the detection target in the current training image frame according to the current training image frame;

[0027] Generate a training target image frame corresponding to each training image frame and containing only the detection target according to the position information of the detection target corresponding to each training image frame;

[0028] For each training target image frame, inputting the current training target image frame into a preset posture recognition model, so that the preset posture recognition model generates training key point information in the current training target image frame according to the current training target image frame;

[0029] The training key point information corresponding to all training target image frames is spliced ​​to generate a training key point sequence corresponding to the power grid monitoring video sample;

[0030] According to the training key point sequence and the behavior judgment label corresponding to the training key point sequence, the first training sample is constructed.

[0031] Furthermore, the training of the target detection model includes:

[0032] Obtaining a number of second training samples;

[0033] Input each second training sample into the target detection model in sequence, and train the target detection model until a preset second training number is reached; wherein, when the target detection model receives a second training sample, it outputs the predicted position information of the detection target corresponding to the second training sample; calculates the second loss function value according to the predicted position information of the detection target and the corresponding image label; and updates the target detection model according to the second loss function value.

[0034] Furthermore, after inputting the key point sequence into the behavior judgment model so that the behavior judgment model judges whether there is sabotage behavior in the key point sequence according to the key point sequence and generates a behavior judgment result, the method further includes:

[0035] In the event of sabotage, an alarm message is generated.

[0036] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0037] An embodiment of the present invention provides a device for identifying sabotage behavior, including: a power grid monitoring video acquisition module, a target image frame generation module, a posture recognition module, a key point sequence generation module and a behavior judgment module;

[0038] The power grid monitoring video acquisition module is used to acquire the power grid monitoring video to be identified;

[0039] The target image frame generation module is used to extract a target image frame containing only the detection target according to the power grid monitoring video to be identified;

[0040] The gesture recognition module is used for inputting the current target image frame into a preset gesture recognition model for each target image frame, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame;

[0041] The key point sequence generation module is used to splice all key point information to generate a key point sequence;

[0042] The behavior judgment module is used to input the key point sequence into the behavior judgment model so that the behavior judgment model judges whether there is human sabotage behavior in the key point sequence based on the key point sequence and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; each first training sample includes a training key point sequence generated by a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

[0043] Based on the above method embodiment, the present invention provides a corresponding electronic device embodiment.

[0044] An embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method for identifying sabotage behavior as described in any one of the above method embodiments can be implemented.

[0045] Based on the above method item embodiments, the present invention provides a corresponding storage medium item embodiment.

[0046] An embodiment of the present invention provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for identifying human-sabotage behavior described in any one of the above method embodiments can be implemented.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The embodiment of the present invention provides a method, device, electronic device and storage medium for identifying sabotage behavior. The method obtains a power grid monitoring video to be identified, performs frame sampling on the video, and generates a plurality of image frames. For each image frame, the image frame is input into a target detection model to generate the position information of the detection target in the image frame, and a target image frame containing only the detection target is generated according to the position information of the detection target in the image frame. Subsequently, the target image frame is input into a posture recognition model to extract the key point information in the image, and the key point information corresponding to all image frames is spliced ​​to form a key point sequence reflecting the target behavior. The key point sequence is input into a behavior judgment model to output the behavior judgment result. This scheme uses the target detection model to accurately extract the target position information, and then generates a key point sequence through posture recognition, effectively captures the dynamic characteristics of the character's behavior, and deeply analyzes the action semantics of the target through the behavior judgment model, thereby distinguishing normal behavior from abnormal sabotage behavior, and improving the accuracy and reliability of recognition. By combining target detection, posture recognition and behavior judgment, automatic recognition of sabotage behavior in power grid monitoring video is realized. By implementing the present invention, the accuracy of sabotage behavior recognition can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The figure is a flow chart of a method for identifying sabotage behavior provided by an embodiment of the present invention.

[0050] Figure 2 The figure is a schematic diagram of the structure of a device for identifying sabotage behavior provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying sabotage behavior, which at least includes the following steps:

[0053] Step S1: Obtain a power grid monitoring video to be identified.

[0054] In the specific implementation, the real-time power grid monitoring video to be identified is collected by high-definition cameras. The advantages of high-definition cameras are not only reflected in image clarity, but also in their ability to work under low light conditions, wide-angle field of view and anti-interference, which enable them to operate stably in various complex environments. These cameras are usually deployed in places where power grid facilities are susceptible to external forces, such as power poles, cables, substation fences, etc., and the monitoring range covers key areas that may be damaged by human intervention.

[0055] By collecting real-time power grid monitoring videos through high-definition cameras, a continuous and complete view of the power grid operation can be obtained, ensuring that relevant image data can be captured in the first time at any time when human sabotage may occur, providing accurate input for subsequent behavioral judgments in a timely manner.

[0056] Step S2: extracting a target image frame containing only the detection target according to the power grid monitoring video to be identified.

[0057] In a preferred embodiment, the step of extracting a target image frame containing only the detection target according to the power grid monitoring video to be identified includes:

[0058] Perform frame sampling on the power grid monitoring video to be identified to generate a number of image frames;

[0059] For each image frame, the current image frame is input into the target detection model, so that the target detection model generates the position information of the detection target in the current image frame according to the current image frame; wherein the target detection model is trained by a plurality of second training samples; each second training sample includes an image frame sample and a picture label corresponding to the image frame sample; the picture label includes the position information of the detection target;

[0060] According to the position information of the detected target in each image frame, a target image frame corresponding to each image frame and containing only the detected target is generated.

[0061] Specifically, the frame sampling of the power grid monitoring video to be identified to generate a plurality of image frames includes:

[0062] According to the preset duration, the power grid monitoring video to be identified is divided into several video clips;

[0063] According to a preset sampling frequency, each video clip is frame sampled to generate a number of original image frames;

[0064] Performing Gaussian denoising on each original image frame to generate a number of denoised image frames;

[0065] Image enhancement is performed on each denoised image frame to generate a plurality of image frames.

[0066] Exemplarily, if the duration of the power grid monitoring video to be identified is 60 minutes, the entire video is divided into several 5-minute video clips according to the preset duration, such as every 5 minutes. In this way, the original 60-minute video will be divided into 12 video clips. The content of each clip is a continuous power grid monitoring screen, and each video clip is frame sampled according to the preset sampling frequency, such as sampling 1 frame per second. Assuming that the duration of each video clip is 5 minutes, each video clip contains 300 seconds of video content. According to the frequency of sampling 1 frame per second, 300 image frames are extracted from each video clip. For 12 video clips, the entire video will generate 3600 original image frames.

[0067] Optionally, a Gaussian filter is used to perform Gaussian denoising on each original image frame to generate several denoised image frames; a contrast adaptive histogram equalization method is used to perform image enhancement on each denoised image frame to improve the brightness and contrast of the image, and several image frames are generated to ensure the accuracy of subsequent detection.

[0068] Optionally, the training of the target detection model includes:

[0069] Obtaining a number of second training samples;

[0070] Input each second training sample into the target detection model in sequence, and train the target detection model until a preset second training number is reached; wherein, when the target detection model receives a second training sample, it outputs the predicted position information of the detection target corresponding to the second training sample; calculates the second loss function value according to the predicted position information of the detection target and the corresponding image label; and updates the target detection model according to the second loss function value.

[0071] Specifically, according to the position information of the detected target in each image frame, a target image frame corresponding to each image frame and containing only the detected target is generated, including: after the target detection model generates the position information of the detected target in each image frame, the position information returned by the target detection model here usually includes the bounding box coordinates of the detected target, such as the coordinates of the upper left corner and the lower right corner, and according to the bounding box coordinates of the detected target, the bounding box of the detected target is annotated using an annotation tool; according to the bounding box of the detected target, the part containing the target area is cropped from the original image, and the target image frame corresponding to each image frame and containing only the detected target is generated. Each cropped image will contain only one target, and the background information of other areas is removed.

[0072] Step S3: for each target image frame, input the current target image frame into a preset gesture recognition model, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame.

[0073] In a specific implementation, the preset posture recognition model is the OpenPose model. OpenPose is a popular posture recognition model that recognizes the posture of people in images through a deep learning algorithm and returns the coordinate information of these joints. The current target image frame posture recognition model is used to detect and extract the key point information of the human body, and output the positions of all key points of the human body in the current image. These key points usually include the positions of various joints of the human body, such as the head, shoulders, elbows, knees, etc. Each key point corresponds to a coordinate, indicating the position of the joint in the image, which can depict a person's posture and movement. The output of each frame of the image is a data containing multiple key point coordinates.

[0074] Step S4: splice all key point information to generate a key point sequence;

[0075] For example, assuming that N key points are detected in a target image frame, the gesture recognition model will output the coordinate information of these key points. For example, the coordinates of the key points of a target image frame may be: [(x1, y1), (x2, y2), ..., (x N ,y N )], where each pair (x, y) represents the position of a key point. For multiple frames, the key point information extracted from each frame is spliced ​​in chronological order to form a key point sequence. A key point sequence is a collection of key point data arranged in chronological order. For example, if three consecutive target image frames are processed, the key point data of each frame will be merged into a long sequence in sequence. Assuming that each target image frame has N key points, the spliced ​​key point sequence contains all the key point coordinates in the multiple frames, which can be expressed as:

[0076] [(x 1_1 ,y 1_1 ),(x 2_1 ,y 2_1 ),...,(x N_1 ,y N_1 ),(x 1_2 ,y 1_2 ),(x 2_2 ,y 2_2 ),...,(x N_2 ,y N_2 ),...,(x 1_t ,y 1_t ),(x 2_t ,y2_t ),...,(x N_t ,y N_t )]

[0077] Here, t represents the total number of frames, and each (x, y) corresponds to the coordinates of a key point in each frame. By splicing, a time series containing all key points can be obtained.

[0078] It can be understood that the spliced ​​key point sequence not only retains the spatial information of the key points in each target image frame, but also connects the key points in multiple frames in time sequence. This sequence can be used for subsequent action analysis and can reflect the behavior dynamics of the character over a period of time.

[0079] Step S5, inputting the key point sequence into the behavior judgment model, so that the behavior judgment model judges whether there is human sabotage behavior in the key point sequence according to the key point sequence, and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; each first training sample includes a training key point sequence generated by a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

[0080] In an optional embodiment, the training of the behavior judgment model includes:

[0081] Obtaining a number of first training samples;

[0082] Input each first training sample into the behavior judgment model in sequence, and train the behavior judgment model until a preset first training number is reached; wherein, the behavior judgment model outputs a predicted behavior judgment result corresponding to the first training sample each time it receives a first training sample; calculates a first loss function value based on the predicted behavior judgment result and the corresponding behavior judgment label; and updates the behavior judgment model based on the first loss function value.

[0083] Preferably, the first training sample is obtained by:

[0084] Obtain power grid monitoring video samples and corresponding behavior judgment labels;

[0085] Perform frame sampling on power grid monitoring video samples to generate several training image frames;

[0086] For each training image frame, input the current training image frame into the target detection model, so that the target detection model generates the position information of the detection target in the current training image frame according to the current training image frame;

[0087] Generate a training target image frame corresponding to each training image frame and containing only the detection target according to the position information of the detection target corresponding to each training image frame;

[0088] For each training target image frame, inputting the current training target image frame into a preset posture recognition model, so that the preset posture recognition model generates training key point information in the current training target image frame according to the current training target image frame;

[0089] The training key point information corresponding to all training target image frames is spliced ​​to generate a training key point sequence corresponding to the power grid monitoring video sample;

[0090] According to the training key point sequence and the behavior judgment label corresponding to the training key point sequence, the first training sample is constructed.

[0091] In a specific implementation, after inputting the key point sequence into the behavior judgment model so that the behavior judgment model judges whether there is sabotage behavior in the key point sequence according to the key point sequence and generates a behavior judgment result, it also includes:

[0092] In the event of sabotage, an alarm message is generated.

[0093] For example, after there is an act of vandalism, an alarm message is generated. The alarm message includes time information and location information. At the same time, the alarm message is quickly pushed to the maintenance personnel through the message queue to ensure that the maintenance personnel receive the alarm in time. The time information is the specific timestamp of the event, which is convenient for subsequent tracking. The location information involves the location of the equipment or area, which helps to locate the location of the event.

[0094] It should be noted that sabotage refers to intentional acts that are carried out by people to damage, destroy or hinder normal operation. Such acts usually target power facilities or equipment, which may cause equipment failure, abnormal system operation, and even endanger the safety and reliability of the entire power grid. Unlike accidents or natural disasters, sabotage is characterized by subjective intent and clear destructiveness. In the power system, sabotage is usually physical damage, which causes physical damage to the equipment by directly acting on it or indirectly interfering with its normal working environment. For example: climbing poles, cutting cables, smashing equipment, and illegally hanging heavy objects on transmission lines.

[0095] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0096] like Figure 2As shown, an embodiment of the present invention provides a device for identifying sabotage behavior, including: a power grid monitoring video acquisition module, a target image frame generation module, a posture recognition module, a key point sequence generation module and a behavior judgment module;

[0097] The power grid monitoring video acquisition module is used to acquire the power grid monitoring video to be identified;

[0098] The target image frame generation module is used to extract a target image frame containing only the detection target according to the power grid monitoring video to be identified;

[0099] The gesture recognition module is used for inputting the current target image frame into a preset gesture recognition model for each target image frame, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame;

[0100] The key point sequence generation module is used to splice all key point information to generate a key point sequence;

[0101] The behavior judgment module is used to input the key point sequence into the behavior judgment model so that the behavior judgment model judges whether there is human sabotage behavior in the key point sequence based on the key point sequence and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; each first training sample includes a training key point sequence generated by a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

[0102] It should be noted that the embodiment of the device described above corresponds to the above-mentioned embodiment of the present invention, and it can implement any one of the above-mentioned methods for identifying acts of vandalism of the present invention. In addition, the embodiment of the above-mentioned device is merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the embodiment of the device provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0103] Based on the above method embodiment of the present invention, a corresponding electronic device embodiment is provided.

[0104] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for identifying destructive behavior described in any one of the present invention is implemented, or when the processor executes the computer program, the functions of each module in the above-mentioned device embodiments are implemented.

[0105] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0106] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0107] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0108] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0109] Based on the above method embodiment, the present invention provides a storage medium embodiment;

[0110] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute any one of the above-mentioned methods for identifying sabotage behaviors of the present invention.

[0111] The storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0112] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0113] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying sabotage behavior, characterized in that: include: Obtain the power grid monitoring video to be identified; Extracting target image frames containing only the detection target according to the power grid monitoring video to be identified; For each target image frame, inputting the current target image frame into a preset gesture recognition model, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame; All key point information is spliced ​​to generate a key point sequence; Inputting the key point sequence into the behavior judgment model, so that the behavior judgment model judges whether there is human-sabotage behavior in the key point sequence according to the key point sequence, and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; Each first training sample includes a training key point sequence generated through a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

2. A method for identifying sabotage behavior as claimed in claim 1, characterized in that: The step of extracting a target image frame containing only the detection target according to the power grid monitoring video to be identified includes: Perform frame sampling on the power grid monitoring video to be identified to generate a number of image frames; For each image frame, the current image frame is input into the target detection model, so that the target detection model generates the position information of the detection target in the current image frame according to the current image frame; wherein the target detection model is trained by a plurality of second training samples; each second training sample includes an image frame sample and a picture label corresponding to the image frame sample; the picture label includes the position information of the detection target; According to the position information of the detected target in each image frame, a target image frame corresponding to each image frame and containing only the detected target is generated.

3. A method for identifying sabotage behavior as claimed in claim 2, characterized in that: The frame sampling of the power grid monitoring video to be identified is performed to generate a plurality of image frames, including: According to the preset duration, the power grid monitoring video to be identified is divided into several video clips; According to a preset sampling frequency, each video clip is frame sampled to generate a number of original image frames; Performing Gaussian denoising on each original image frame to generate a number of denoised image frames; Image enhancement is performed on each denoised image frame to generate a plurality of image frames.

4. A method for identifying sabotage behavior as claimed in claim 3, characterized in that: The training of the behavior judgment model includes: Obtaining a number of first training samples; Input each first training sample into the behavior judgment model in sequence, and train the behavior judgment model until a preset first training number is reached; wherein, the behavior judgment model outputs a predicted behavior judgment result corresponding to the first training sample each time it receives a first training sample; calculates a first loss function value based on the predicted behavior judgment result and the corresponding behavior judgment label; and updates the behavior judgment model based on the first loss function value.

5. A method for identifying sabotage behavior as claimed in claim 4, characterized in that: The first training sample is obtained by: Obtain power grid monitoring video samples and corresponding behavior judgment labels; Perform frame sampling on power grid monitoring video samples to generate several training image frames; For each training image frame, input the current training image frame into the target detection model, so that the target detection model generates the position information of the detection target in the current training image frame according to the current training image frame; Generate a training target image frame corresponding to each training image frame and containing only the detection target according to the position information of the detection target corresponding to each training image frame; For each training target image frame, inputting the current training target image frame into a preset posture recognition model, so that the preset posture recognition model generates training key point information in the current training target image frame according to the current training target image frame; The training key point information corresponding to all training target image frames is spliced ​​to generate a training key point sequence corresponding to the power grid monitoring video sample; According to the training key point sequence and the behavior judgment label corresponding to the training key point sequence, the first training sample is constructed.

6. A method for identifying sabotage behavior as claimed in claim 5, characterized in that: The training of the target detection model includes: Obtaining a number of second training samples; Input each second training sample into the target detection model in sequence, and train the target detection model until a preset second training number is reached; wherein, when the target detection model receives a second training sample, it outputs the predicted position information of the detection target corresponding to the second training sample; calculates the second loss function value according to the predicted position information of the detection target and the corresponding image label; and updates the target detection model according to the second loss function value.

7. A method for identifying sabotage behavior as claimed in claim 6, characterized in that: After inputting the key point sequence into the behavior judgment model so that the behavior judgment model judges whether there is sabotage behavior in the key point sequence according to the key point sequence and generates a behavior judgment result, the method further includes: In the event of sabotage, an alarm message is generated.

8. A device for identifying sabotage behavior, characterized in that: include: Power grid monitoring video acquisition module, target image frame generation module, posture recognition module, key point sequence generation module and behavior judgment module; The power grid monitoring video acquisition module is used to acquire the power grid monitoring video to be identified; The target image frame generation module is used to extract a target image frame containing only the detection target according to the power grid monitoring video to be identified; The gesture recognition module is used for inputting the current target image frame into a preset gesture recognition model for each target image frame, so that the preset gesture recognition model generates key point information in the current target image frame according to the current target image frame; The key point sequence generation module is used to splice all key point information to generate a key point sequence; The behavior judgment module is used to input the key point sequence into the behavior judgment model, so that the behavior judgment model judges whether there is human-sabotage behavior in the key point sequence according to the key point sequence, and generates a behavior judgment result; wherein the behavior judgment model is trained by a plurality of first training samples; Each first training sample includes a training key point sequence generated through a power grid monitoring video sample, and a behavior judgment label corresponding to the training key point sequence for identifying whether there is human sabotage behavior in the power grid monitoring video sample.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for identifying sabotage behavior described in any one of claims 1 to 7 can be implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement a method for identifying sabotage behavior as described in any one of claims 1 to 7.

Images