Power grid safety monitoring method based on abnormal behavior perception

Abstract

The invention provides a power grid safety monitoring method based on abnormal behavior perception, relates to the field of image recognition and detection, and is suitable for state monitoring and safety management of power equipment in a complex construction environment. The abnormal behavior perception-based power grid safety monitoring method specifically comprises the following steps of: acquiring a construction site image at the current moment by using image acquisition equipment, and detecting to obtain position information of transportation equipment; acquiring state information according to the detected transportation equipment; acquiring dust concentration information generated by the transportation equipment; and determining the influence degree of the power equipment according to the position information, the state information and the flying dust concentration information of the transportation equipment, and performing early warning. Through the steps, the abnormal condition that the transportation equipment in the construction site generates dust is detected in real time, the potential influence of the dust on the power equipment is analyzed, and early warning is given out in time.

Description

Technical Field

[0001] The present invention relates to the field of image recognition and detection, and in particular to a power grid security monitoring method based on abnormal behavior perception, which is suitable for status monitoring and safety management of power equipment in complex construction environments, effectively warning of potential threats, and ensuring the safe and stable operation of the power grid. Background Art

[0002] With the rapid development of modern infrastructure construction, the scale of various construction sites continues to expand, and the acceleration of urbanization has also brought more complex construction needs. From bridge construction to rail transit construction, from high-rise building construction to underground pipeline laying, all walks of life are inseparable from power supply. As the basic energy for the development of modern society, the rational application and safety management of electricity at the construction site are particularly important. The supervision and maintenance of power equipment at the construction site and the scientific and reasonable construction of the power grid are not only the key to ensuring the smooth progress of the construction project, but also an important prerequisite for ensuring the safety of construction personnel and equipment. Modern construction sites often have complex environments, tight construction periods, and frequent construction activities. The operating status of power equipment is affected by many factors, which makes the management of the power grid face great challenges. Reasonable power grid construction and scientific power equipment maintenance can not only ensure the stability of power consumption during the construction process, but also effectively reduce the risk of safety accidents caused by power problems. Therefore, power grid safety monitoring has gradually become an important part of construction safety management. However, with the continuous expansion of construction scale and increasingly complex site conditions, traditional power grid management methods have gradually failed to meet the needs of modern construction, and more intelligent and refined monitoring methods are urgently needed to deal with power safety issues at construction sites.

[0003] In the prior art, in order to reduce the damage to power equipment at the construction site, some conventional safety protection measures are usually taken. For example, sound and light warning signs are set up around the construction area to remind construction personnel and equipment operators to pay attention to the existence of power equipment, so as to avoid damage to power equipment or electric shock accidents caused by misoperation or negligence. In addition, infrared equipment is used to monitor the activities of construction personnel and mechanical equipment in real time to ensure that the construction operation remains within the safe range of the power equipment. These measures reduce the probability of damage to power equipment caused by human factors at the construction site to a certain extent, and also provide basic safety guarantees for construction personnel. However, with the development of modern construction technology and the increasing complexity of the construction environment, traditional sound and light warning and infrared monitoring methods have gradually exposed some shortcomings. First, these methods mainly focus on the distance monitoring between construction personnel and equipment and power equipment, and lack effective identification capabilities for more potential power grid security threats in the construction site environment; second, these methods usually rely on the vigilance and active prevention awareness of construction personnel, and it is often difficult to timely discover and deal with some hidden safety hazards caused by high concealment or long-term accumulation. Especially in some large construction sites, multiple construction teams and mechanical equipment are working at the same time, and the environment changes frequently, and the limitations of this monitoring method are more prominent.

[0004] In addition, the working environment of the construction site itself also has a significant impact on the operating status of the power equipment. Long-term excavation operations, stone loading and unloading, vehicle transportation and other activities at the construction site will cause the dust concentration in the air to be significantly higher than that in ordinary areas. These dusts diffuse in the air, which will not only affect the health of construction workers, but also accumulate on the surface of power equipment due to electrostatic adsorption. Over time, the dust adsorbed on the surface of power equipment will gradually accumulate, bringing a series of adverse effects on the normal operation of the equipment. What is more complicated is that such problems are often hidden and long-term, and are difficult to be discovered in time, and traditional power grid monitoring methods at construction sites do not provide effective solutions for such problems. As a major characteristic of the construction site, the impact of dust problems on the safe operation of power equipment cannot be ignored. A targeted monitoring method is urgently needed to achieve all-round monitoring and management of dust generation, diffusion and its impact on power equipment. Summary of the invention

[0005] In response to the problems existing in the above-mentioned prior art, the present invention proposes a power grid security monitoring method based on abnormal behavior perception, which performs real-time detection of abnormal conditions of dust generated by various construction equipment at the construction site, analyzes the potential impact of dust on power equipment, and issues early warnings in a timely manner.

[0006] The present invention provides a power grid security monitoring method based on abnormal behavior perception, which specifically comprises the following steps:

[0007] S1: Use image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment;

[0008] S2: Acquiring status information of the transport equipment according to the detection;

[0009] S3: Obtaining dust concentration information generated by the transportation equipment;

[0010] S4: Determine the impact on the power equipment and issue an early warning based on the location information, status information and dust concentration information of the transportation equipment.

[0011] The present invention provides a power grid security monitoring system based on abnormal behavior perception, the system comprising:

[0012] Image acquisition equipment: Use the image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment;

[0013] Status detection module: the status detection module detects and obtains status information of the transport equipment;

[0014] Dust concentration detection module: the dust concentration detection module obtains dust concentration information generated by the transportation equipment;

[0015] An early warning module determines the impact on the power equipment and issues an early warning based on the location information, status information and dust concentration information of the transportation equipment.

[0016] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned power grid security monitoring method based on abnormal behavior perception when executing the computer program.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the above-mentioned power grid security monitoring method based on abnormal behavior perception.

[0018] Compared with the prior art, the present invention aims to perceive the degree of abnormal dust concentration generated by transportation equipment in complex construction sites on power. The present invention evaluates the degree of influence of transportation equipment position, transportation equipment status, power equipment position and dust concentration information on power equipment. Before the evaluation, the transportation equipment position is detected using a multi-scale feature map. Based on the detected transportation equipment position, the transportation equipment status is quickly obtained by using the position migration before and after, key point changes and transportation equipment position expansion area to avoid detecting invalid areas and causing waste of computing resources and time consumption. At the same time, the dust concentration generated by the transportation equipment is regressed and calculated using image optical features, the dust diffusion factors under different states are analyzed, and targeted dust diffusion models are established based on different state information, so as to perceive the abnormal dust generated by the transportation equipment and evaluate its impact on power equipment, thereby realizing supervision and management of safety hazards that are highly concealed and caused by long-term accumulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is a flow chart of the power grid security monitoring based on abnormal behavior perception according to the present invention; DETAILED DESCRIPTION

[0021] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0022] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0023] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspect described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0024] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the examples can be practiced without these specific details.

[0025] The embodiment of this specification proposes a power grid security monitoring method based on abnormal behavior perception, which specifically includes the following steps:

[0026] S1: Use image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment;

[0027] S2: Acquiring status information of the transport equipment according to the detection;

[0028] S3: Obtaining dust concentration information generated by the transportation equipment;

[0029] S4: Determine the impact on the power equipment and issue an early warning based on the location information, status information and dust concentration information of the transportation equipment.

[0030] In one embodiment, the present invention uses multiple industrial-grade high-definition network cameras to capture images of the construction site, and the image acquisition equipment should have high-resolution imaging capabilities to ensure that clear equipment details and scene information can be captured in a complex construction environment, and meet the requirements of subsequent position detection for image accuracy. At the same time, the equipment should have a wide dynamic range (WDR) function to adapt to the complex lighting conditions of strong light and alternating shadows at the construction site, and prevent the image quality from being degraded due to overexposure or underexposure. In addition, the equipment should have good low-light performance and infrared fill light functions to ensure that high-quality images can still be stably collected at night or in an environment with insufficient light. Taking into account the characteristics of dust, high humidity, vibration and large temperature difference in the construction site environment, the image acquisition equipment must also have a high degree of protection, including waterproof, dustproof and shockproof performance, and the protection level should reach IP67 or above to ensure that the equipment can operate stably in a harsh environment for a long time. In addition, the equipment should support edge computing interfaces or have basic intelligent analysis functions, and can directly complete basic data processing (such as target detection, background denoising, etc.) at the acquisition end to reduce the pressure of data transmission. Considering the dynamic changes of the construction site and the installation and maintenance requirements of the equipment, the image acquisition equipment should also support remote configuration and management functions to facilitate dynamic adjustment of equipment parameters according to the progress of the on-site construction. Preferably, the image acquisition equipment can be Hikvision DS-2CD2T87G2-L or Dahua DH-IPC-HFW7842H-Z4, two industrial-grade high-definition network cameras.

[0031] The installation layout of the image acquisition equipment should be flexibly designed according to the area, terrain, range of movement of the transportation equipment and possible obstruction factors of the construction site. The design principles include: installing several fixed cameras in the surrounding areas of the construction site to ensure full coverage of the entire construction site. The installation height of the camera is set to 5 meters to 8 meters according to the characteristics of the specific area. In high-activity density areas such as passages and stacking areas where transportation equipment frequently enters and exits, the camera installation height is set to 8 meters to expand the field of view; in relatively less active areas, the height is set to 5 meters to ensure that the specific details of the transportation equipment can be captured. The horizontal viewing angle of each camera is set to 90°, the vertical viewing angle is set to 45°, and the overlapping area of ​​the field of view between adjacent equipment is controlled at 15%-20%, which not only avoids the occurrence of monitoring blind spots, but also ensures that other cameras can achieve compensatory monitoring when the equipment is blocked or fails.

[0032] In response to the environmental impacts such as dust and water vapor that may exist at the construction site, all cameras are equipped with automatic optical cleaning devices, including lens heating and automatic wiping functions, to prevent dust or water droplets from adhering to the lens surface and causing blurred images. In addition, to improve the durability of the equipment, each camera is also equipped with an anti-vibration bracket to ensure that the equipment will not shift or be damaged due to the impact force generated by ground vibration or mechanical operation during construction. In order to cope with the large changes in light at the construction site, each camera is equipped with an infrared fill light with a fill light distance of 50 meters, which can perform high-definition shooting at night or in low-light environments.

[0033] The collected image data is transmitted to the edge computing device through a combination of wired network (optical fiber) and wireless network (5G). Multiple edge computing units are deployed at the construction site. Preferably, the core hardware can use the NVIDIA Jetson AGX Orin module, which has a computing power of up to 200TOPS and can support real-time image processing and analysis functions. The edge computing device is responsible for preprocessing the image data collected on site, including denoising, compression, target detection and other operations. The processed key data and metadata are uploaded to the cloud server through the 5G network. The cloud server is responsible for the storage and global analysis of large-scale data. The edge device and the cloud achieve two-way communication through the MQTT protocol. The cloud analysis results can be fed back to the edge in real time to guide subsequent inspections and scheduling on site.

[0034] In order to meet the dynamic needs of the construction site, the acquisition system adopts a zoning management and dynamic adjustment strategy. The construction site is divided into several monitoring areas, and at least two fixed cameras are deployed in each area to achieve cross coverage. When the monitoring needs of a certain area change due to construction progress or equipment movement, the system can automatically adjust the camera's shooting direction, focal length and parameter settings through the central control platform to ensure that the scene can always be fully captured. The central control platform also has an equipment status diagnosis function, which can monitor the operating status of all image acquisition devices in real time. When a device fails, it will automatically trigger the alarm mechanism and dispatch spare equipment for replacement.

[0035] The transport equipment location information is obtained through a transport equipment detection network, which includes two shallow feature extraction modules: and Four deep feature extraction modules Feature enhancement network and object detection head;

[0036] The two shallow feature extraction modules and Feature extraction by 3*3 convolution:

[0037]

[0038] in, and Respectively represent the input and output of the first or second shallow feature extraction module, where the input of the first shallow feature extraction module is the input image, and the input of the second shallow feature extraction module is the output feature map of the first shallow feature module. Conv, BN and σ represent convolution, normalization and activation operations, respectively;

[0039] In the deep feature extraction module Defined as:

[0040]

[0041] in, Respectively represent the input feature map, intermediate output feature map and final output feature map of the first or second deep feature extraction module, Fuse 1 represents a feature fusion operation, is the first or second enhanced feature map of the feature enhancement network, d and e have the same value each time, and the feature fusion operation is defined as:

[0042]

[0043] Among them, feat1 and feat2 are two feature maps to be fused. represents pixel-by-pixel feature fusion;

[0044] In the deep feature extraction module Defined as:

[0045]

[0046] in, Respectively represent the input feature map and the final output feature map of the third or fourth deep feature extraction module, Fuse 2 represents the secondary feature fusion operation, It represents the third or fourth enhanced feature map of the feature enhancement network, d and e have the same value each time, and the secondary feature fusion operation is defined as:

[0047]

[0048] Among them, λ is the fusion coefficient constant;

[0049] The feature enhancement network is used to retain shallow feature information into deep features for guiding target detection. The feature enhancement network is defined as:

[0050]

[0051] in, The first to fourth enhanced feature maps of the feature enhancement network;

[0052] The target detection head is defined as:

[0053]

[0054]

[0055] in, are the six output feature maps of the feature enhancement network, and is the i-th intermediate feature map and output feature map of the target detection head, represents element-by-element multiplication fusion, represents the output feature map with a position annotation box, UpSml, FC represent the upsampling and fully connected layers respectively, and the position annotation box marks the center position of the target and the length and width information of the target area;

[0057] When transport equipment is detected in a construction site environment, due to the complexity of the construction site environment, there are many kinds of mechanical equipment and personnel, and there are unfavorable factors such as mutual occlusion and background interference in the scene, making accurate detection of transport equipment extremely challenging. The present invention effectively solves the problem of transport equipment detection in complex scenes by introducing a technology that integrates shallow features and deep features. Shallow features focus on extracting detailed information of target objects, such as edge contours, texture features, and local structures, which can well reflect the detailed features of the target; while deep features have higher semantic expression capabilities and can extract high-level semantic information in complex backgrounds, such as the category of transport equipment, overall shape features, etc. However, in complex construction scenes, the single use of deep features often leads to a decrease in the recognition ability of the target equipment due to the lack of detailed information, especially in the presence of background interference or when the target is far away. To this end, the present invention innovatively deeply fuses the rich detailed information contained in the shallow features with the semantic information expressed in the deep features, and realizes the organic combination of detailed features and semantic features in the detection process, thereby greatly improving the detection and recognition accuracy of transport equipment in complex background environments. Through this feature fusion method, the present invention can better adapt to adverse conditions such as light changes, equipment occlusion, and complex backgrounds at the construction site, and provides strong technical support for the accurate detection of transportation equipment.

[0058] In addition, the present invention also adopts multi-scale feature fusion detection technology to deal with the problem that the parking position of transportation equipment at the construction site is not fixed and the target presents different scales in the image. In the detection of transportation equipment at the construction site, the size of the target in the image varies greatly because the equipment may be at different distances and viewing angles. If these multi-scale characteristics are not fully considered, the performance of the detection model may be significantly reduced, especially when detecting distant equipment, missed detection or false detection may occur. The present invention integrates feature information at different scales through the fusion of multi-scale features, so that the detection model can take into account the feature expression of large and small targets at the same time, thereby effectively improving the detection ability of targets of different scales. For transportation equipment close to the camera, the model can capture its large-scale features and accurately extract the contour and semantic information of the equipment; while for transportation equipment far away from the camera, the model can use small-scale features to capture the key details of distant targets to ensure the accuracy and completeness of the detection. Multi-scale feature fusion not only improves the detection ability of close and distant targets, but also can maintain stable detection performance when the target size changes greatly, greatly enhancing the adaptability and practicality of the present invention in complex construction scenarios.

[0059] Acquiring status information of the transport equipment according to the detection, wherein the status information includes four status information: driving, waiting, dumping and loading;

[0060] S2-1: Acquire the construction site image at the next moment and detect the location information of the same transportation equipment as that in the construction site image at the current moment;

[0061] S2-2 calculates the intersection-and-joint ratio of the position information of the same transport equipment at the previous and next moments, and when the intersection-and-joint ratio is less than a preset moving threshold, feedback is given that the transport equipment is in a driving state, otherwise, step S2-3 is executed;

[0062] S2-3: Use the key point detection model to perform key point detection on the same transport equipment at the previous and next moments, and obtain the dumping plane according to the key points. If the dumping plane at the previous and next moments is greater than the preset dumping threshold, the transport equipment is fed back as being in a dumping state, otherwise, execute step S2-4;

[0063] S2-4: Set the area to which the transportation equipment belongs in the construction site image at the current moment as an invalid mask area, and set an extended mask area at the same time, wherein the extended mask area includes the invalid mask area, and detect heavy equipment based on the invalid mask area and the extended mask area. When heavy equipment is detected, feedback is given that the transportation equipment is in the shipping state, otherwise feedback is given that the transportation equipment is in the waiting state.

[0064] Before monitoring whether the dust at the construction site interferes with the power equipment for abnormal perception, the present invention first classifies the operating status of the source of dust - the transportation equipment, including driving, loading, dumping and waiting states. This is a key step in achieving accurate judgment of the impact of dust. The construction site environment is complex, and the source of dust is not only related to the location of the transportation equipment, but also significantly affected by the current state of the equipment. Under different states, there are significant differences in the intensity of dust generation and the range of dust diffusion of transportation equipment. For example, in the driving state, the moving speed of the equipment will directly affect the intensity and diffusion direction of the dust, especially under the action of wind, the dust will form a significant deviation with the driving direction of the equipment and the wind direction; in the shipping state, the loading and unloading of materials may cause large local dust. At this time, the diffusion range of the dust is relatively concentrated but the intensity is high, and the diffusion direction is mostly related to the wind direction; the dumping state is often accompanied by the rapid unloading of a large amount of materials, which causes the intensity of the dust source to increase instantly, and the diffusion range is also closely related to the dumping direction; in the waiting state, the equipment is in a stationary state and the dust intensity is relatively low. By classifying and detecting the operating status of the transportation equipment, the present invention can effectively distinguish the dust characteristics under different states, and then establish a targeted dust diffusion model according to the current state of the equipment.

[0065] This dust modeling method based on state classification can improve the accuracy of judging the impact of abnormal dust. In practical applications, the impact of transportation equipment in different states on dust has different spatial distribution characteristics. If a unified diffusion model is used without state distinction, it will often lead to misjudgment of the dust impact range. For example, for the driving state, if the effect of the equipment's driving direction and wind direction is not fully considered, the dust diffusion range may be underestimated; for the dumping state, if the dumping direction is not considered, the intensity of the dust's impact on the surrounding power equipment may be misjudged. Therefore, before the abnormal perception, the present invention can realize the dynamic classification management of dust sources by accurately detecting the operating state of the transportation equipment, ensuring that the dust diffusion model in different states can truly reflect the actual situation on site. This method not only improves the accuracy of supervision and management, but also provides a scientific basis for the perception and response of subsequent abnormal dust interference.

[0066] When the transport equipment is in a driving state, the driving direction and speed of the transport equipment are obtained by calculating the coordinate offset of the central area position of the transport equipment at the previous and next moments;

[0067] The key point detection model is used to detect key points of the same transport equipment at the previous and next moments. Specifically, the detected transport equipment area image is input into the initial feature extraction network to obtain the initial feature map F0, and the initial feature map is input into the multi-resolution path network. The multi-resolution path network includes three stages Stage1-Stage3, and the paths of the three stages are 2, 3, 3, respectively.

[0068] The first stage, Stage 1, defines two paths:

[0069] The second stage Stage2 defines three paths:

[0070] The third stage Stage3 defines three paths:

[0071] A cross-fusion module is set between the first stage and the second stage and between the second stage and the third stage, and the cross-fusion module is defined as:

[0072]

[0073] in, represents the mth path feature graph in the kth stage, Indicates that the resolution of the mth path feature map is adjusted to the resolution of the nth path feature map in the kth stage, argetRes=n indicates that the target resolution is the resolution of the nth path feature map, k is 1 or 2, and the values ​​of m and n are determined according to the number of paths in the stage;

[0074] According to the third stage, the highest resolution path Generate heatmaps and extract key points;

[0075]

[0076] (x r ,y r ) = argmax (x,y) HeatMap r ,r∈{1,2,3,4};

[0077] Among them, HeatMap r is the rth heat map, is the rth key point extracted at the current moment, and Corresponding to the left and right corners of the front of the dump bucket, and There are two left and right corner points at the rear of the dump box.

[0078] The dumping plane is obtained according to the key points. When the inclination angle of the dumping plane is greater than the preset dumping threshold at the previous and next moments, the feedback that the transport equipment is in the dumping state specifically includes:

[0079] The current moment key point and the next key point Substitute r∈{1,2,3,4} into the plane equation: z=ax+by+c, and get the current tipping plane z t =a t x+b t y+c t and the next moment tipping bucket plane z t+1 =a t+1 x+b t+1 y+c t+1 , calculate the inclination angle of the dump plane before and after Where n t and n t+1 are the normal vectors of the dump bucket plane at the current moment and the next moment respectively, and the normal vector direction at the current moment is taken as the dumping direction;

[0080] The area to which the transport equipment belongs in the current construction site image is set as an invalid mask area, and an extended mask area is set at the same time, wherein the extended mask area includes the invalid mask area, and heavy equipment is detected based on the invalid mask area and the extended mask area. When heavy equipment is detected, feedback is given that the transport equipment is in a shipping state, otherwise feedback is given that the transport equipment is in a waiting state. Specifically, the following steps are performed:

[0081] Get the center position of the transport equipment location information x ,center y ) and the length and width information (width, height) of the target area, setting the extended pixels of the extended area to d>0, indicating the distance extended outward based on the boundary box of the transport equipment;

[0082] Invalid mask area:

[0083]

[0084] Extended mask area:

[0085]

[0086] Valid mask area:

[0087] M eff (x,y)=M ext (x,y)-M inv (x,y);

[0088] The effective mask area is superimposed on the current construction site image to obtain an effective detection area image, and based on the effective detection area image, a target detection model is used to determine whether there is heavy equipment in the effective detection area image.

[0089] Obtaining the dust concentration information generated by the transportation equipment specifically includes:

[0090] Calculate the optical image features of the transportation equipment area in the construction site image at the current moment, and estimate the dust concentration value using a dust concentration regression model based on the optical image features;

[0091] The optical image features include brightness mean, brightness standard deviation, image blur, edge density and grayscale difference;

[0092] The brightness mean is:

[0093]

[0094] The brightness standard deviation is:

[0095]

[0096] The image blur:

[0097]

[0098] The edge density:

[0099]

[0100] The grayscale difference:

[0101]

[0102] Among them, I g (x, y) is the image of the transportation equipment area in the construction site image at the current moment, N E is the number of edges detected by the Canny operator, and P(i,j) is the probability value of the gray-level co-occurrence matrix;

[0103] Among them, the training of dust concentration regression model includes:

[0104] Dataset collection and annotation: Collect image data of the transportation equipment area of ​​the construction site to ensure coverage of different weather, lighting conditions and dust concentration ranges. Each image is synchronized with the corresponding real dust concentration value (unit: μg / m^3) through a dust sensor (such as a PM10 sensor) as a training label.

[0105] Data preprocessing: The training samples are cropped to the transportation equipment area, and grayscale conversion and data enhancement are performed. After processing, the brightness mean, brightness standard deviation, image blur, edge magic, and grayscale difference of the transportation equipment area are extracted as the optical image feature values ​​of each image.

[0106] Training process: The data set is divided into a training set and a test set at an 8:2 ratio. The second-order polynomial regression model is initialized, and the mean square error (MSE) is constructed as the loss function. During the training process, the optical image feature values ​​of each image are input into the regression model, and the model parameters are optimized using gradient descent. The hyperparameters are adjusted through grid search or random search until the loss function converges or the set number of iterations is reached.

[0107] Determining the impact of the power equipment and issuing an early warning based on the location information, status information and dust concentration information of the transportation equipment includes:

[0108] Construct dust dispersion models for driving, dumping and loading states respectively:

[0109] Dust dispersion model:

[0110]

[0111] Among them, C state is the dust source intensity, and σ τ are the standard deviations of the diffusion direction and dust diffusion, respectively, reflecting the dust diffusion range, and τ(x,y) is the diffusion dust concentration value at the position with coordinates (x,y);

[0112] When the transport equipment is in motion:

[0113] C state =C*v*ω1;

[0114]

[0115] Among them, d drive is the total diffusion direction of the driving state, which is the vector sum of the driving direction and the wind direction. and is the component of the total diffusion direction in the driving state along the x direction and the y direction, v represents the driving speed, ω1 is the driving dust intensity coefficient, and C is the dust concentration value;

[0116] When the transport equipment is in the dumping and shipping state, C state = C, where, when the transport equipment is in a dumped state, d pour is the total diffusion direction of the dumping state, which is the vector sum of the dumping direction and the wind direction. and It is the component of the total diffusion direction in the dumping state along the x direction and the y direction; when the transport equipment is in the dumping state: d wind is the total diffusion direction of the shipping state, which is also the wind direction vector, and is the component of the total diffusion direction in the shipping state along the x-direction and the y-direction;

[0117] When the transport device is in the waiting state, set τ(x,y) to 0;

[0118] The power equipment location information is substituted into the dust diffusion model, and the diffusion dust concentration value at the power equipment location is estimated based on the detected transportation equipment acquisition status information and wind direction information, and an early warning message is issued when the diffusion dust concentration value is greater than a preset threshold.

[0119] The present invention provides a power grid security monitoring system based on abnormal behavior perception, the system comprising:

[0120] Image acquisition equipment: Use the image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment;

[0121] Status detection module: the status detection module detects and obtains status information of the transport equipment;

[0122] Dust concentration detection module: the dust concentration detection module obtains dust concentration information generated by the transportation equipment;

[0123] An early warning module determines the impact on the power equipment and issues an early warning based on the location information, status information and dust concentration information of the transportation equipment.

[0124] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned power grid security monitoring method based on abnormal behavior perception when executing the computer program.

[0125] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the above-mentioned power grid security monitoring method based on abnormal behavior perception.

[0126] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0127] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0128] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A power grid security monitoring method based on abnormal behavior perception, characterized in that: The method comprises the following steps: S1: Use image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment; S2: Acquiring status information of the transport equipment according to the detection; S3: Obtaining dust concentration information generated by the transportation equipment; S4: Determine the impact on the power equipment and issue an early warning based on the location information, status information and dust concentration information of the transportation equipment.

2. A power grid security monitoring method based on abnormal behavior perception according to claim 1, characterized in that: The transport equipment location information is obtained through a transport equipment detection network, which includes two shallow feature extraction modules. and Four deep feature extraction modules Feature enhancement network and object detection head.

3. A power grid security monitoring method based on abnormal behavior perception according to claim 1, characterized in that: Acquiring status information of the transport equipment according to the detection, wherein the status information includes four status information: driving, waiting, dumping and loading; S2-1: Acquire the construction site image at the next moment and detect the location information of the same transportation equipment as that in the construction site image at the current moment; S2-2 calculates the intersection-and-joint ratio of the position information of the same transport equipment at the previous and next moments. When the intersection-and-joint ratio is less than a preset moving threshold, the transport equipment is fed back to be in a driving state. Otherwise, step S2-3 is executed. S2-3: Use the key point detection model to perform key point detection on the same transport equipment at the previous and next moments, and obtain the dumping plane according to the key points. When the inclination angle of the dumping plane at the previous and next moments is greater than the preset dumping threshold, feedback is given that the transport equipment is in a dumping state, otherwise, execute step S2-4; S2-4: Set the area to which the transportation equipment belongs in the construction site image at the current moment as an invalid mask area, and set an extended mask area at the same time, wherein the extended mask area includes the invalid mask area, and detect heavy equipment based on the invalid mask area and the extended mask area. When heavy equipment is detected, feedback is given that the transportation equipment is in the shipping state, otherwise feedback is given that the transportation equipment is in the waiting state.

4. The power grid security monitoring method based on abnormal behavior perception according to claim 3 is characterized by: The key point detection model is used to detect key points of the same transport equipment at the previous and next moments. Specifically, the detected transport equipment area image is input into the initial feature extraction network to obtain the initial feature map F0, and the initial feature map is input into the multi-resolution path network. The multi-resolution path network includes three stages Stage1-Stage3, and the paths of the three stages are 2, 3, 3, respectively. The first stage, Stage 1, defines two paths: The second stage, Stage 2, defines three paths: The third stage Stage3 defines three paths: A cross-fusion module is set between the first stage and the second stage and between the second stage and the third stage, and the cross-fusion module is defined as: in, represents the mth path feature graph in the kth stage, Indicates that the resolution of the mth path feature map is adjusted to the resolution of the nth path feature map in the kth stage, argetRes=n indicates that the target resolution is the resolution of the nth path feature map, k is 1 or 2, and the values ​​of m and n are determined according to the number of paths in the stage; According to the third stage, the highest resolution path Generate heatmap and extract key points; (x r ,y r )=argmax (x,y) HeatMap r ,r∈{1,2,3,4}; Among them, HeatMap r is the rth heat map, is the rth key point extracted at the current moment, and Corresponding to the left and right corners of the front of the dump bucket, and There are two left and right corner points at the rear of the dump box.

5. The power grid security monitoring method based on abnormal behavior perception according to claim 4 is characterized by: The dumping plane is obtained according to the key points. When the inclination angle of the dumping plane is greater than the preset dumping threshold at the previous and next moments, the feedback that the transport equipment is in the dumping state specifically includes: The current moment key point and the next key point Substitute r∈{1,2,3,4} into the plane equation: z=ax+by+c, where a, b, c are the plane equation parameters, and we get the current tipping plane z t =a t x+b t y+c t and the next moment tipping bucket plane z t+1 =a t+1 x+b t+1 y+c t+1 , calculate the inclination angle of the dump plane before and after where n t and n t+1 are the normal vectors of the dump bucket plane at the current moment and the next moment respectively, and the direction of the normal vector at the current moment is taken as the dumping direction.

6. The power grid security monitoring method based on abnormal behavior perception according to claim 5 is characterized by: Calculate the optical image features of the transportation equipment area in the construction site image at the current moment, and estimate the dust concentration value using a dust concentration regression model based on the optical image features; The optical image features include brightness mean, brightness standard deviation, image blur, edge density and grayscale difference.

7. A power grid security monitoring method based on abnormal behavior perception according to claim 6, characterized in that: Determining the impact of the power equipment and issuing an early warning based on the location information, status information and dust concentration information of the transportation equipment includes: Construct dust dispersion models for driving, dumping and loading states respectively: Among them, C state is the dust source intensity, and σ τ are the standard deviations of the diffusion direction and dust diffusion, respectively, reflecting the dust diffusion range, and τ(x,y) is the diffusion dust concentration value at the position with coordinates (x,y); When the transport equipment is in motion: C state =C*v*ω1; Among them, d drive is the total diffusion direction of the driving state, which is the vector sum of the driving direction and the wind direction. and is the component of the total diffusion direction of the driving state along the x direction and the y direction, v represents the driving speed, ω1 is the driving dust intensity coefficient, and C is the dust concentration value; When the transport equipment is in the dumping and shipping state, C state = C, where, when the transport equipment is in a dumped state, d pour is the total diffusion direction of the dumping state, which is the vector sum of the dumping direction and the wind direction. and It is the component of the total diffusion direction in the dumping state along the x direction and the y direction; when the transport equipment is in the dumping state: d wind is the total diffusion direction of the shipping state, which is also the wind direction vector, and is the component of the total diffusion direction of the shipping state in the x and y directions; when the transport equipment is in the waiting state, set τ(x,y) to 0; The power equipment location information is substituted into the dust diffusion model, and the diffusion dust concentration value at the power equipment location is estimated based on the detected transportation equipment acquisition status information and wind direction information, and an early warning message is issued when the diffusion dust concentration value is greater than a preset threshold.

8. A power grid security monitoring system based on abnormal behavior perception, used to execute a power grid security monitoring method based on abnormal behavior perception as claimed in any one of claims 1 to 7, characterized in that: The system includes: Image acquisition equipment: Use the image acquisition equipment to obtain the current construction site image and detect the location information of the transportation equipment; Status detection module: the status detection module detects and obtains status information of the transport equipment; Dust concentration detection module: the dust concentration detection module obtains dust concentration information generated by the transportation equipment; An early warning module determines the impact on the power equipment and issues an early warning based on the location information, status information and dust concentration information of the transportation equipment.

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 monitoring power grid security based on abnormal behavior perception as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements a power grid security monitoring method based on abnormal behavior perception as claimed in any one of claims 1 to 7.

Images