Power distribution construction safety distance monitoring method, device and equipment and storage medium

Abstract

The invention provides a power distribution construction safety distance monitoring method, device and equipment and a storage medium, and the method comprises the steps: inputting an obtained power distribution operation scene image into a stereo matching neural network model for binocular stereo matching calculation, outputting a three-dimensional information disparity map, and converting the three-dimensional information disparity map into a power distribution operation depth map; based on the power distribution operation scene image and the target detection neural network model, key operation elements and two-dimensional coordinate information are determined; screening the depth values of the pixel points in the key job element target frame to obtain image coordinates of the screened pixel points; performing coordinate system conversion on the image coordinates of the screened pixel points and the depth values to obtain initial three-dimensional point cloud data of the key job elements; clustering and screening the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; and determining a first nearest point coordinate, a second nearest point coordinate and a third nearest point coordinate from the effective three-dimensional point cloud data, and determining a safety distance among the operator, the construction machinery and the electrified equipment.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of safety monitoring for power construction operations, including but not limited to a method, device, equipment, and storage medium for monitoring the safety distance in distribution construction. Background Art

[0002] During the construction, renovation, and maintenance of the power transmission and distribution system, especially when installing key equipment such as distribution transformers in mountainous areas with complex terrain and limited working space, construction safety faces a severe test. Operations usually involve the coordinated operation of large construction machinery, on-site cooperation of multiple construction workers, and the possible presence of adjacent energized lines or equipment. Ensuring sufficient safety distances between various operation units, between personnel and construction machinery, and between personnel and energized equipment is the core prerequisite for preventing safety accidents such as electric shock and collision and ensuring personal and equipment safety.

[0003] Currently, the monitoring of safety distances at distribution construction sites mainly relies on manual visual inspection and experience judgment. The on-site safety officers or construction workers rely on experience and visual observation to judge the distances between objects. This method is highly subjective and is easily affected by various factors such as the observer's state (such as fatigue, distraction), weather conditions (such as rain, fog, and light changes), and observation angles. It is difficult to guarantee the accuracy and cannot perform continuous and dynamic monitoring.

[0004] However, in outdoor scenarios such as mountainous distribution construction, which are open and have a complex background, how to efficiently and accurately integrate object detection and three-dimensional depth perception, precisely extract the independent point clouds of multiple target objects from the raw data containing a large amount of background noise and interference, and calculate the minimum spatial distance between them remains a challenging technical problem. Therefore, it is of great practical significance to develop an intelligent method that can overcome the defects of traditional methods and achieve real-time and accurate three-dimensional positioning and safety distance monitoring of multiple key targets at distribution construction operation sites. Summary of the Invention

[0005] Based on the problems in the related art, embodiments of the present invention provide a method, device, equipment, and storage medium for monitoring the safety distance in distribution construction.

[0006] The technical solution of the embodiments of the present invention is implemented as follows:

[0007] Embodiments of the present invention provide a method for monitoring the safety distance in distribution construction, the method including:

[0008] Obtaining a distribution operation scene image of a distribution operation site by using a binocular vision image acquisition device;

[0009] Input the power distribution operation scene image into a pre - constructed stereo matching neural network model for binocular stereo matching calculation, and output a three - dimensional information disparity map; based on the binocular camera calibration parameters, convert the three - dimensional information disparity map into a power distribution operation depth map;

[0010] Based on the power distribution operation scene image and a pre - constructed target detection neural network model, determine key operation elements and two - dimensional coordinate information; at least the key operation elements include operating personnel, construction machinery, and energized equipment;

[0011] Based on the power distribution operation depth map, the two - dimensional coordinate information, and preset depth threshold information, screen the depth values of the pixel points within the target box of the key operation elements to obtain the image coordinates of the screened pixel points;

[0012] Convert the image coordinates of the screened pixel points and the depth values into a coordinate system to obtain the initial three - dimensional point cloud data of the key operation elements; perform clustering screening on the initial three - dimensional point cloud data to obtain effective three - dimensional point cloud data;

[0013] Using a preset K - D tree algorithm, determine the first nearest point coordinates between the operating personnel point cloud and the construction machinery point cloud, the second nearest point coordinates between the operating personnel point cloud and the energized equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the energized equipment point cloud from the effective three - dimensional point cloud data;

[0014] Based on the first nearest point coordinates, the second nearest point coordinates, and the third nearest point coordinates, determine the safety distances between the operating personnel, the construction machinery, and the energized equipment.

[0015] An embodiment of the present invention provides a power distribution construction safety distance monitoring device, and the device includes:

[0016] An acquisition module, configured to acquire a power distribution operation scene image of a power distribution operation site by using a binocular vision image acquisition device;

[0017] A calculation module, configured to input the power distribution operation scene image into a pre - constructed stereo matching neural network model for binocular stereo matching calculation, and output a three - dimensional information disparity map; based on the binocular camera calibration parameters, convert the three - dimensional information disparity map into a power distribution operation depth map;

[0018] A determination module, configured to determine key operation elements and two - dimensional coordinate information based on the power distribution operation scene image and a pre - constructed target detection neural network model; at least the key operation elements include operating personnel, construction machinery, and energized equipment;

[0019] A screening module, configured to screen the depth values of the pixel points within the target box of the key operation elements based on the power distribution operation depth map, the two-dimensional coordinate information, and the preset depth threshold information, so as to obtain the image coordinates of the screened pixel points;

[0020] A conversion module, configured to perform coordinate system conversion on the image coordinates of the screened pixel points and the depth values to obtain the initial three-dimensional point cloud data of the key operation elements; perform clustering screening on the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data;

[0021] The determination module is further configured to use a preset K-D tree algorithm to determine the first nearest point coordinates between the operator point cloud and the construction machinery point cloud, the second nearest point coordinates between the operator point cloud and the energized equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the energized equipment point cloud from the effective three-dimensional point cloud data;

[0022] The determination module is further configured to determine the safety distance between the operator, the construction machinery, and the energized equipment based on the first nearest point coordinates, the second nearest point coordinates, and the third nearest point coordinates.

[0023] In some embodiments, the pre-constructed stereo matching neural network model is the IGEV++ model; the calculation module is further configured to adopt a multi-range strategy, calculate the feature correlation volume group by group in the small disparity range, and perform sparse calculation of the feature correlation volume in the large disparity range through adaptive patch matching to determine the matching degree at different disparities; use a lightweight 3D convolutional network model to regularize the multi-range cost volume, fuse the multi-scale geometric context information to obtain the aggregated cost volume; perform weighted summation on the aggregated cost volume through a Soft-argmin operation to calculate the continuous disparity values and generate an initial disparity map; perform iterative update based on ConvGRU, combine a selective feature fusion module, and use multi-range geometric features to optimize and correct the initial disparity map to obtain the three-dimensional information disparity map.

[0024] In some embodiments, the determination module is further configured to input the power distribution operation scene image into a pre-constructed target detection neural network model for recognition and analysis to locate the key operation elements; output the target box of the key operation elements and the two-dimensional coordinate information thereof in the image coordinate system; use the binocular camera calibration parameters to convert the disparity values in the three-dimensional information disparity map into the power distribution operation depth map including the depth values of each point in the power distribution operation site; where the conversion formula is:

[0025] ;

[0026] In the formula, is the depth value of each point in the depth map of the power distribution operation; is the focal length of the binocular vision image acquisition device; is the horizontal baseline distance of the binocular vision image acquisition device; is the disparity value.

[0027] In some embodiments, the pre-constructed target detection neural network model is the D-FINE model. The network uses HGNetv2 as the feature extraction backbone structure, and connects the encoder-decoder of the Transformer architecture to process features of different scales. Finally, the model prediction results are output through the traditional regression head and the D-FINE head.

[0028] In some embodiments, the coordinate system conversion formula is:

[0029] ;

[0030] In the formula, and are the focal lengths of the camera in the , directions respectively; ([[]] ,, ) is the image coordinate of the filtered pixel point; ([[]] ,, ) is the central coordinate of the imaging plane of the binocular camera; R and T are the external parameter rotation matrix and offset matrix of the camera; ([[]] ,, , ) is the initial three-dimensional point cloud data.

[0031] In some embodiments, the clustering and screening algorithm is the Mean Shift algorithm. By iteratively calculating the density-weighted mean within the neighborhood of each point, the points are gradually moved to the local density maximum points. Finally, the points converging to the same peak are grouped into one cluster, and the cluster structure can be automatically recognized.

[0032] In some embodiments, the first nearest point coordinate is , ; the second nearest point coordinate is , ; the third nearest point coordinate is , ;

[0033] The safety distance calculation formula is:

[0034] ;

[0035] ;

[0036] ;

[0037] Among them, d1 represents the Euclidean distance between the operator's point cloud and the construction machinery's point cloud, d2 represents the Euclidean distance between the operator's point cloud and the energized equipment's point cloud, and d3 represents the Euclidean distance between the construction machinery's point cloud and the energized equipment's point cloud.

[0038] An embodiment of the present invention provides a distribution construction safety distance monitoring device, including: a memory for storing executable instructions; a processor for implementing the above-mentioned distribution construction safety distance monitoring method when executing the executable instructions stored in the memory.

[0039] An embodiment of the present invention provides a computer-readable storage medium storing executable instructions for causing a processor to implement the above-mentioned distribution construction safety distance monitoring method when executing the executable instructions.

[0040] The distribution construction safety distance monitoring method, device, equipment and storage medium provided by the embodiments of the present invention obtain high-precision scene disparity information by combining a pre-constructed binocular stereo matching network model, and use the target detection neural network model D-FINE to accurately identify various operation elements and their positions in the image, and then construct a depth map of the distribution operation site based on the binocular camera parameters; further, by performing depth screening on the depth map, converting it into a point cloud and separating the operation element point cloud from the complex on-site point cloud through the MeanShift clustering algorithm, effectively removing noise and environmental interference; finally, using the K-D tree to efficiently search for and determine the nearest points between the operator's point cloud, the construction machinery's point cloud, and the energized equipment's point cloud, and accurately calculating the Euclidean distances between the personnel and the energized equipment and the construction machinery, so as to realize the automatic, high-precision, and real-time monitoring of the distribution construction safety distance. Especially when performing distribution construction operations under complex terrain conditions such as mountains, the present invention can overcome the influence of factors such as the environment, light, and personnel's subjective judgment of traditional methods, provide stable and reliable safety distance information, significantly improve the intrinsic safety level during the power construction process, effectively prevent the occurrence of safety accidents such as electric shock, and ensure the life safety of construction personnel. Description of the Drawings

[0041] Figure 1 is a schematic structural diagram of a distribution construction safety distance monitoring system provided by an embodiment of the present invention;

[0042] Figure 2 is a schematic flow diagram of a distribution construction safety distance monitoring method provided by an embodiment of the present invention;

[0043] Figure 3 is a schematic flow diagram of a distribution construction safety distance monitoring method based on binocular vision and point cloud clustering provided by an embodiment of the present invention;

[0044] Figure 4 It is a schematic structural diagram of the IGEV++ stereo matching network model provided by an embodiment of the present invention;

[0045] Figure 5 It is a schematic composition structure diagram of the power distribution construction safety distance monitoring device provided by an embodiment of the present invention;

[0046] Figure 6 It is a schematic composition structure diagram of the electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0047] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be construed as limiting the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0048] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used in the embodiments of the present invention have the same meaning as commonly understood by those skilled in the technical field to which the embodiments of the present invention belong. The terms used in the embodiments of the present invention are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0049] The following describes an exemplary application of the power distribution construction safety distance monitoring device according to an embodiment of the present invention. The power distribution construction safety distance monitoring device provided by an embodiment of the present invention can be implemented as a terminal or as a server. In one implementation, the power distribution construction safety distance monitoring device provided by an embodiment of the present invention can be implemented as various types of terminals such as a laptop computer, a tablet computer, a desktop computer, a mobile device, etc.; in another implementation, the power distribution construction safety distance monitoring device provided by an embodiment of the present invention can also be implemented as a server. Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery network (CDN, Content Delivery Network), and big data and artificial intelligence platforms. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, which are not limited in the embodiments of the present invention. Below, an exemplary application when the power distribution construction safety distance monitoring device is implemented as a server will be described.

[0050] SeeFigure 1 , Figure 1 is a schematic structural diagram of a power distribution construction safety distance monitoring system 10 provided by an embodiment of the present invention. An embodiment of the present invention can provide a power distribution construction safety distance monitoring platform, and the power distribution construction safety distance monitoring platform can be implemented as a power distribution construction safety distance monitoring application. The power distribution construction safety distance monitoring system 10 provided by an embodiment of the present invention includes a terminal 110, a network 120, and a server 130. Among them, the server 130 is the server of the power distribution construction safety distance monitoring application. The server 130 can constitute the power distribution construction safety distance monitoring device of the embodiment of the present invention. The terminal 110 is connected to the server 130 through the network 120, and the network 120 can be a wide area network, a local area network, or a combination of the two.

[0051] In some embodiments, please refer to Figure 1 , when monitoring the safety distance of a power distribution construction site, the terminal 110 sends the initiated power distribution construction safety distance monitoring task to the server 130 through the network 120. In response to the power distribution construction safety distance monitoring task sent by the terminal 110, the server 130 uses a binocular vision image acquisition device to obtain a power distribution operation scene image of the power distribution operation site; inputs the power distribution operation scene image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation, and outputs a three-dimensional information disparity map; based on the binocular camera calibration parameters, converts the three-dimensional information disparity map into a power distribution operation depth map; based on the power distribution operation scene image and a pre-constructed target detection neural network model, determines key operation elements and two-dimensional coordinate information; based on the power distribution operation depth map, two-dimensional coordinate information, and a preset depth threshold information, filters the depth values of the pixel points within the target box of the key operation elements to obtain the image coordinates of the filtered pixel points; performs coordinate system conversion on the image coordinates and depth values of the filtered pixel points to obtain the initial three-dimensional point cloud data of the key operation elements; performs clustering filtering on the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; uses a preset K-D tree algorithm to determine the first nearest point coordinates between the operator point cloud and the construction machinery point cloud, the second nearest point coordinates between the operator point cloud and the live equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the live equipment point cloud from the effective three-dimensional point cloud data; based on the first nearest point coordinates, second nearest point coordinates, and third nearest point coordinates, determines the safety distance between the operator, construction machinery, and live equipment. After obtaining the safety distance, the server 130 sends the safety distance to the terminal 110 through the network 120.

[0052] It should be noted here that cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or local area network to achieve data computing, storage, processing, and sharing. Cloud technology is the general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model. It can form a resource pool, be used on demand, and is flexible and convenient. Cloud computing technology will become an important support. The back-end services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites, and more portal websites. With the high development and application of the Internet industry, in the future, each item may have its own identification mark and needs to be transmitted to the back-end system for logical processing. Data at different levels will be processed separately, and various industry data requires the support of a powerful system back-end, which can only be achieved through cloud computing.

[0053] In some embodiments, there may also be a cloud memory, and the stereo matching neural network model pre-constructed by the server can be stored in the cloud memory. Then, when the server obtains the power distribution operation scene image, it can perform binocular stereo matching calculations based on the stored stereo matching neural network model to obtain a three-dimensional information disparity map.

[0054] In some embodiments, there may also be a cloud memory, and the target detection neural network model pre-constructed by the server can be stored in the cloud memory. Then, when the server obtains the power distribution operation scene image, it can perform identification and analysis based on the stored target detection neural network model to locate key operation elements.

[0055] An embodiment of the present invention provides a method for monitoring the safety distance in power distribution construction. Refer to Figure 2 , Figure 2 which is a schematic flowchart of a method for monitoring the safety distance in power distribution construction provided by an embodiment of the present invention, and will be described in conjunction with Figure 2 the steps shown.

[0056] Step S210, use a binocular vision image acquisition device to obtain a power distribution operation scene image of the power distribution operation site.

[0057] In some embodiments, the binocular vision image acquisition device is composed of two cameras with a fixed limit distance (i.e., the horizontal distance between the optical centers of the two cameras). At the power distribution operation site, this device can simultaneously obtain two images with disparities from different visual shooting scenes.

[0058] In some embodiments, the power distribution operation scenario image refers to the image data collected by a binocular vision image acquisition device at the power distribution operation site, which contains information such as the environment of the operation site, the operation behaviors of the operators, the placement and operating states of various construction machinery, and the working conditions of energized equipment.

[0059] In the present invention, the power distribution operation scenario image includes a left image of the power distribution operation scenario and a right image of the power distribution operation scenario.

[0060] Step S220: Input the power distribution operation scenario image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation, and output a three-dimensional information disparity map; based on the binocular camera calibration parameters, convert the three-dimensional information disparity map into a power distribution operation depth map.

[0061] In some embodiments, the pre-constructed stereo matching neural network model refers to a neural network model constructed based on deep learning technology, which is specifically used to process the left and right perspective images obtained by a binocular vision image acquisition device. Its core function is to calculate the disparity of each pixel point in the left and right images by matching the features of the same object in the two images. This model usually consists of multiple convolutional layers, pooling layers, and fully connected layers. During the training stage, a large number of binocular image data with labeled disparity information are used, and the model parameters are continuously adjusted through the backpropagation algorithm, so that it can accurately perform stereo matching calculation on the newly input images and output a three-dimensional information disparity map.

[0062] In some embodiments, the three-dimensional information disparity map refers to the result output after processing the power distribution operation scenario image by the stereo matching neural network model, which is an image with the same size as the original image. In the disparity map, the value of each pixel point represents the disparity of this point in the left and right images, that is, the position difference in the horizontal direction. Here, the larger the disparity value, the closer the object corresponding to this point is to the camera; the smaller the value, the farther the object is from the camera. The disparity map intuitively reflects the distance relationship of the objects in the scene and provides key data for subsequent conversion into a depth map.

[0063] In some embodiments, the binocular camera calibration parameters refer to the data set obtained by measuring the parameters of the two cameras in the binocular vision image acquisition device through a specific calibration method. It includes internal parameters and external parameters. The internal parameters are used to describe the imaging characteristics of a single camera, such as focal length, principal point coordinates, distortion coefficients, etc., which reflect the conversion relationship between projecting points in three-dimensional space onto the two-dimensional image plane by the camera; the external parameters are used to describe the relative position and attitude relationship between the two cameras, including the rotation matrix and the translation vector. These parameters are necessary conditions for converting the three-dimensional information disparity map into a power distribution operation depth map. Only by accurately calibrating the camera parameters can the accurate conversion from image pixel coordinates to actual scene space coordinates be achieved.

[0064] In some embodiments, the depth map of the power distribution operation refers to an image obtained through certain mathematical calculations based on the three-dimensional information disparity map and the calibration parameters of the binocular camera. The value of each pixel in the depth map represents the actual distance of that point from the camera in the actual power distribution operation scenario, usually in meters or millimeters. The depth map is presented in the form of grayscale or color, and different grayscale values or colors correspond to different depths, which can intuitively show the position distribution of various objects in the three-dimensional space at the operation site, providing intuitive data support for subsequent in-depth analysis of key operation elements.

[0065] Step S230: Based on the power distribution operation scenario image and the pre-constructed target detection neural network model, determine the key operation elements and two-dimensional coordinate information; at least the key operation elements include operators, construction machinery, and energized equipment.

[0066] In some embodiments, the target detection neural network model refers to a neural network model based on deep learning, which is mainly used to identify and locate key operation elements in the power distribution operation scenario image. By learning a large number of image data annotated with various key operation elements (such as operators, construction machinery, energized equipment, etc.), this model extracts the feature patterns of these elements, so as to be able to detect the corresponding targets in the newly input image and determine their two-dimensional coordinate information in the image. Common model structures include YOLO, Faster R-CNN, D-FINE model, etc. By continuously optimizing the network structure and training parameters, the detection accuracy and speed of key operation elements in different sizes, poses, and environments are improved.

[0067] In some embodiments, the key operation elements refer to the objects or entities that have an important impact on operation safety and operation procedures during the power distribution operation. Among them, the operator is the person directly involved in the power distribution operation, and their behavior and position are crucial for operation safety; construction machinery such as cranes and aerial work platforms are important tools for completing the power distribution operation tasks, and their operating status and position also affect operation safety; the energized equipment is in an energized operating state, and it is crucial to ensure operation safety that the operator and construction machinery maintain a safe distance from the energized equipment.

[0068] In some embodiments, the two-dimensional coordinate information refers to the position information of the target element on the image plane determined by the target detection neural network model after detecting the key operation elements in the power distribution operation scenario image. It is usually represented in the form of a rectangular box, and the position and size of the target in the image are determined by the upper left corner coordinates (x1, y1) and the lower right corner coordinates (x2, y2) of the rectangular box.

[0069] Step S240: Based on the power distribution operation depth map, the two-dimensional coordinate information, and the preset depth threshold information, screen the depth values of the pixel points within the target box of the key operation elements to obtain the image coordinates of the screened pixel points.

[0070] In some embodiments, the preset depth threshold information refers to the range standard preset for screening the depth values of key operation elements according to factors such as the safety specifications of power distribution operations, the characteristics of the operation environment, and the performance of equipment. For example, considering the safety protection distance requirements of live equipment, a minimum depth threshold is set. Pixel points below this threshold may indicate that the object is too close and there are safety risks. At the same time, a maximum depth threshold is set. Pixel points exceeding this threshold may not be included in the key analysis scope due to reasons such as too far distance or measurement errors. By using the preset depth threshold information, pixel points that meet the actual analysis requirements can be effectively screened, improving the accuracy and efficiency of subsequent calculations.

[0071] Step S250: Perform coordinate system conversion on the image coordinates of the screened pixel points and the depth values to obtain the initial three-dimensional point cloud data of the key operation elements; perform clustering screening on the initial three-dimensional point cloud data to obtain the effective three-dimensional point cloud data.

[0072] In some embodiments, the initial three-dimensional point cloud data refers to the data set obtained by performing coordinate system conversion on the image coordinates of the screened pixel points and the corresponding depth values.

[0073] In some embodiments, the effective three-dimensional point cloud data refers to the data obtained by performing clustering screening on the initial three-dimensional point cloud data. During the clustering screening process, clustering algorithms (such as the density-based clustering algorithm DBSCAN, the K-Means clustering algorithm, etc.) are used to analyze the initial point cloud data. According to the distance and density relationship between points, the densely distributed points are divided into different categories, removing isolated noise points and outliers, and retaining the effective point cloud data that can truly reflect the shape and position characteristics of the key operation elements.

[0074] Step S260: Use the preset K-D tree algorithm to determine the first nearest point coordinates between the operator point cloud and the construction machinery point cloud, the second nearest point coordinates between the operator point cloud and the live equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the live equipment point cloud from the effective three-dimensional point cloud data.

[0075] In some embodiments, the K-D tree (K-Dimensional tree) algorithm is a data structure and algorithm for efficiently organizing and searching data points in a K-dimensional space. In this solution, the preset K-D tree algorithm is used to quickly and accurately search for and determine the nearest point coordinates between different key operation element point clouds from the effective three-dimensional point cloud data. By continuously dividing the three-dimensional space and constructing a tree structure, this algorithm can quickly exclude most irrelevant regions during the nearest neighbor search, greatly improving the search efficiency. It is especially suitable for processing large-scale three-dimensional point cloud data and can find the nearest point coordinates between the operator, construction machinery, and energized equipment point clouds in a short time.

[0076] Step S270: Based on the first nearest point coordinate, the second nearest point coordinate, and the third nearest point coordinate, determine the safety distance between the operator, the construction machinery, and the energized equipment.

[0077] In some embodiments, the safety distance refers to the minimum distance that needs to be maintained between the operator, construction machinery, and energized equipment as stipulated according to the safety codes and standards of distribution operations.

[0078] In practical applications, different types of energized equipment, different operating environments, and operating requirements correspond to different safety distance standards. By calculating the actual distance between key operation elements and comparing it with the safety distance, it can be determined whether the distribution operation is in a safe state. If the actual distance is less than the safety distance, a safety warning is issued to remind the operator to take corresponding safety measures to ensure the safety of the operation.

[0079] The power distribution construction safety distance monitoring method, device, equipment and storage medium provided by the embodiments of the present invention obtain high-precision scene parallax information by combining a pre-constructed binocular stereo matching network model, and use the target detection neural network model D-FINE to accurately identify various operation elements and their positions in the image, and then construct a depth map of the power distribution operation site based on the binocular camera parameters; further, by performing depth screening on the depth map, converting it into a point cloud and separating the operation element point cloud from the complex on-site point cloud through the MeanShift clustering algorithm, effectively removing noise and environmental interference; finally, using the K-D tree to efficiently search and determine the closest points between the operator point cloud and the construction machinery point cloud and the live equipment point cloud, and accurately calculating the Euclidean distances between the personnel and the live equipment and the construction machinery, so as to realize the automatic, high-precision and real-time monitoring of the power distribution construction safety distance. Especially when performing power distribution construction operations under complex terrain conditions such as mountainous areas, the present invention can overcome the influence of factors such as environment, light and human subjective judgment of traditional methods, provide stable and reliable safety distance information, significantly improve the intrinsic safety level during the power construction process, effectively prevent the occurrence of safety accidents such as electric shock, and ensure the life safety of construction personnel.

[0080] In some embodiments, the pre-constructed stereo matching neural network model is the IGEV++ model; the above step S220 can be implemented through the following steps S221 to S224:

[0081] Step S221, adopting a multi-range strategy, calculating the feature correlation volume group by group in the small parallax range, and sparsely calculating the feature correlation volume by adaptive patch matching in the large parallax range to determine the matching degree at different parallaxes.

[0082] Step S222, using a lightweight 3D convolutional network model to regularize the multi-range cost volume, fuse multi-scale geometric context information, and obtain the aggregated cost volume.

[0083] Step S223, performing weighted summation on the aggregated cost volume through the Soft-argmin operation, calculating continuous parallax values, and generating an initial parallax map.

[0084] Step S224, based on ConvGRU for iterative update, combining a selective feature fusion module, and using multi-range geometric features to optimize and correct the initial parallax map to obtain the three-dimensional information parallax map.

[0085] In some embodiments, "determining the key operation elements and two-dimensional coordinate information based on the distribution operation scene image and the pre-constructed target detection neural network model" in the above step S230 is implemented as follows: inputting the distribution operation scene image into the pre-constructed target detection neural network model for identification and analysis to locate the key operation elements; and outputting the target box of the key operation elements and the two-dimensional coordinate information thereof in the image coordinate system.

[0086] In some embodiments, "converting the three-dimensional information disparity map into a distribution operation depth map based on the binocular camera calibration parameters" in the above step S230 is implemented as follows: using the binocular camera calibration parameters to convert the disparity values in the three-dimensional information disparity map into the distribution operation depth map containing the depth values of each point in the distribution operation site; where the conversion formula is:

[0087] ;

[0088] In the formula, is the depth value of each point in the distribution operation depth map; is the focal length of the binocular vision image acquisition device; is the horizontal baseline distance of the binocular vision image acquisition device; is the disparity value.

[0089] In some embodiments, the pre-constructed target detection neural network model in the above step S230 is the D-FINE model. The network uses HGNetv2 as the feature extraction backbone structure, connects the encoder-decoder of the Transformer architecture to process features of different scales, and finally outputs the model prediction results through the traditional regression head and the D-FINE head.

[0090] In some embodiments, the coordinate transformation formula in the above step S250 is:

[0091] ;

[0092] In the formula, and are the focal lengths of the camera in the , directions respectively; ([[]] , , ) is the image coordinate of the selected pixel point; ([[]] , , ) is the central coordinate of the imaging surface of the binocular camera; R and T are the external parameter rotation matrix and offset matrix of the camera; ([[]] , , , ) is the initial three-dimensional point cloud data.

[0093] In some embodiments, the clustering and screening algorithm is the Mean Shift algorithm. By iteratively calculating the density-weighted mean within the neighborhood of each point, the points are gradually moved to the local density maximum points, and finally, the points converging to the same peak are grouped into one cluster, which can automatically identify the cluster structure.

[0094] In some embodiments, the coordinates of the first nearest point are , ; the coordinates of the second nearest point are , ; the coordinates of the third nearest point are , ;

[0095] The safety distance calculation formula is:

[0096] ;

[0097] ;

[0098] ;

[0099] where d1 represents the Euclidean distance between the point cloud of the operator and the point cloud of the construction machinery, d2 represents the Euclidean distance between the point cloud of the operator and the point cloud of the energized equipment, and d3 represents the Euclidean distance between the point cloud of the construction machinery and the point cloud of the energized equipment.

[0100] Next, the exemplary application of the embodiments of the present invention in an actual application scenario will be described.

[0101] The present invention provides a method for monitoring the safety distance of distribution construction based on binocular vision and point cloud clustering. As Figure 3 shown, it includes the following steps:

[0102] Step 1: Use a binocular camera to collect the left and right images of the personnel distribution operation scene, and construct a distribution construction data set containing various operation elements.

[0103] Step 2: Input the left and right images of the collected data set into the stereo matching neural network model, perform binocular stereo matching calculation, and output a disparity map reflecting the three-dimensional information of each point in the scene.

[0104] Among them, the stereo matching neural network model is specifically the IGEV++ model. As Figure 4 shown, it specifically includes the following steps:

[0105] Step 201: Matching cost calculation: Adopt a multi-range strategy. Calculate the feature correlation volume group by group in the small disparity range, and calculate the correlation volume sparsely through adaptive patch matching in the large disparity range to capture the matching degree at different disparities.

[0106] Step 202, cost aggregation: Regularize the multi-range cost volume using a lightweight 3D convolutional network (3D UNet), fuse multi-scale geometric context information, and optimize the cost volume.

[0107] Step 203, disparity calculation: Perform weighted summation on the aggregated cost volume through the Soft-argmin operation, calculate continuous disparity values, and generate an initial disparity map.

[0108] Step 204, disparity optimization: Perform iterative updates using ConvGRU, combine with a selective feature fusion module, and optimize and correct the initial disparity map using multi-range geometric features to improve disparity accuracy.

[0109] Step 3, Input the left image in the dataset into the target detection neural network model. The model analyzes the image, identifies and locates the key operation elements in the image, such as operators, construction machinery, and energized equipment, and outputs the target boxes of these elements and their precise coordinate information in the image coordinate system; among them, the target detection neural network model is specifically the D-FINE model. The network uses HGNetv2 as the feature extraction backbone structure, connects the encoder-decoder of the Transformer architecture to process features of different scales, and finally outputs the model prediction results through the traditional regression head and the D-FINE head.

[0110] Step 4, Use the calibrated binocular camera parameters to convert the disparity map obtained in Step 2 into a depth map containing the depth information of each point in the scene; among them, the camera calibration uses the Zhang Zhengyou calibration method to analyze the internal and external parameters of the camera and the distortion coefficients. Convert the disparity value D in the disparity map into a depth map containing the depth value Z of each point in the scene. The formula is as follows:

[0111] ;

[0112] In the formula, is the depth value of each point in the distribution operation depth map; is the focal length of the binocular vision image acquisition device; is the horizontal baseline distance of the binocular vision image acquisition device; is the disparity value.

[0113] Step 5, Combine the image coordinate information of each operation element obtained by target detection in Step 3. In the depth map obtained in Step 4, set the depth threshold range, perform validity screening on the depth values of the pixel points within the box, exclude invalid or noise points caused by factors such as occlusion, too far or too close distance, and retain the relatively reliable depth information points of the target object.

[0114] Step 6: Combine the image coordinates of the pixel points within the target selected in Step 5 and the corresponding depth values, and with the binocular camera parameters, convert the two-dimensional image coordinates and depth information into three-dimensional space coordinates through coordinate system transformation, thereby obtaining the three-dimensional point cloud data of the operation object. The formula for converting the two-dimensional image coordinates and depth information into three-dimensional space coordinates (i.e., the initial three-dimensional point cloud data in the above embodiments) is as follows:

[0115] ;

[0116] In the formula, and are the focal lengths of the camera in the , directions respectively; ([[]] , ) are the image coordinates of the selected pixel points; ([[]] , ) are the central coordinates of the imaging plane of the binocular camera; R and T are the external camera parameter rotation matrix and offset matrix; ([[]] , , ) are the initial three-dimensional point cloud data.

[0117] Step 7: For the three-dimensional point cloud data of the operation site obtained in Step 6, perform clustering analysis on it, identify and remove the point clouds of environmental factors that are not directly related to safety distance monitoring, such as background, ground, vegetation, etc., as well as possible noise point clouds, so as to accurately extract the effective point clouds belonging to key operation elements such as operators, construction machinery, and energized equipment. Among them, the point cloud clustering analysis is specifically the Mean Shift algorithm. By iteratively calculating the density-weighted mean (drift vector) within the neighborhood of each point, the points are gradually moved to the local density maximum points, and finally the points converging to the same peak are grouped into one cluster, which can automatically identify the cluster structure.

[0118] Step 8: Use the K-Dimensional Tree algorithm to efficiently search for and determine the nearest points between the operator point cloud and the construction machinery point cloud and the energized equipment point cloud in the selected operation element point clouds, and calculate the Euclidean distance between these nearest point pairs. This distance is the actual safety distance between the operator and the construction machinery and between the operator and the energized equipment at the current moment. Compare this actual distance with the preset safety threshold to achieve real-time monitoring and early warning of the safety distance in power distribution construction. Among them, the coordinates of the nearest points between the operator point cloud and the construction machinery point cloud are , respectively, the coordinates of the nearest points between the operator point cloud and the energized equipment point cloud are , respectively, and the coordinates of the nearest points between the construction machinery point cloud and the energized equipment point cloud are , , the formula for calculating its Euclidean distance d is as follows:

[0119] ;

[0120] ;

[0121] ;

[0122] Among them, d1 represents the Euclidean distance between the point cloud of the operator and the point cloud of the construction machinery, d2 represents the Euclidean distance between the point cloud of the operator and the point cloud of the energized equipment, and d3 represents the Euclidean distance between the point cloud of the construction machinery and the point cloud of the energized equipment.

[0123] Figure 5 is a schematic structural diagram of the composition of the distribution construction safety distance monitoring device provided by an embodiment of the present invention. As Figure 5 shown, the distribution construction safety distance monitoring device 500 includes: an acquisition module 501, configured to acquire a distribution operation scene image of the distribution operation site by using a binocular vision image acquisition device; a calculation module 502, configured to input the distribution operation scene image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation, and output a three-dimensional information disparity map; based on the binocular camera calibration parameters, convert the three-dimensional information disparity map into a distribution operation depth map; a determination module 503, configured to determine key operation elements and two-dimensional coordinate information based on the distribution operation scene image and a pre-constructed target detection neural network model; at least the key operation elements include an operator, construction machinery, and energized equipment; a screening module 504, configured to screen the depth values of the pixel points within the target frame of the key operation elements based on the distribution operation depth map, the two-dimensional coordinate information, and preset depth threshold information, and obtain the image coordinates of the screened pixel points; a conversion module 505, configured to perform coordinate conversion on the image coordinates of the screened pixel points and the depth values to obtain the initial three-dimensional point cloud data of the key operation elements; perform clustering screening on the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; the determination module 503 is further configured to use a preset K-D tree algorithm to determine the first nearest point coordinates between the point cloud of the operator and the point cloud of the construction machinery, the second nearest point coordinates between the point cloud of the operator and the point cloud of the energized equipment, and the third nearest point coordinates between the point cloud of the construction machinery and the point cloud of the energized equipment from the effective three-dimensional point cloud data; the determination module 503 is further configured to determine the safety distance between the operator, the construction machinery, and the energized equipment based on the first nearest point coordinates, the second nearest point coordinates, and the third nearest point coordinates.

[0124] It should be noted that the description of the device according to the embodiments of the present invention is similar to the description of the above method embodiments, and has similar beneficial effects to the method embodiments. Therefore, no further details will be given. For the technical details not disclosed in the embodiments of the present device, please refer to the description of the method embodiments of the present invention for understanding.

[0125] It should be noted that in the embodiments of the present invention, if the above-mentioned method for monitoring the safe distance in power distribution construction is implemented in the form of software functional modules and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the related technology, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a terminal to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM), magnetic disks, or optical discs that can store program codes. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0126] Correspondingly, the embodiments of the present invention provide a device for monitoring the safe distance in power distribution construction. Figure 6 It is a schematic diagram of the composition structure of the device for monitoring the safe distance in power distribution construction provided by the embodiments of the present invention. As Figure 6 shown, the device 600 for monitoring the safe distance in power distribution construction at least includes: a processor 601 and a computer-readable storage medium 602 configured to store executable instructions, where the processor 601 generally controls the overall operation of the device 600 for monitoring the safe distance in power distribution construction. The computer-readable storage medium 602 is configured to store instructions and applications executable by the processor 601, and can also cache data to be processed or already processed by the processor 601 and each module in the device 600 for monitoring the safe distance in power distribution construction, and can be implemented by flash memory (FLASH) or random access memory (RAM).

[0127] The embodiments of the present invention provide a storage medium storing executable instructions, where the executable instructions are stored. When the executable instructions are executed by a processor, the processor will be caused to execute the methods provided by the embodiments of the present invention. For example, as Figure 2 shown in the method.

[0128] In some embodiments, the storage medium may be a computer-readable storage medium. For example, it can be a ferroelectric random access memory (FRAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disc, or compact disk-read only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above memories.

[0129] In some embodiments, the executable instructions may be in the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and can be deployed in any form, including being deployed as an independent program or being deployed as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0130] As an example, the executable instructions may or may not correspond to a file in the file system. They can be stored as part of a file that stores other programs or data. For example, they can be stored in one or more scripts in a hypertext markup language (HTML) document, stored in a single file dedicated to the program in question, or stored in multiple cooperating files (such as files that store one or more modules, subroutines, or code portions). As an example, the executable instructions can be deployed to execute on one electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed at multiple locations and interconnected by a communication network.

[0131] As described above, the above are only embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and scope of the present invention are all included in the protection scope of the present invention.

[0132] It should be understood that the "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present invention. Therefore, the "in one embodiment" or "in an embodiment" that appears throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present invention, the magnitude of the serial numbers of the above processes does not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention. The serial numbers of the embodiments of the present invention above are only for description and do not represent the advantages or disadvantages of the embodiments.

[0133] It should be noted that in this article, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method or device. Without further limitation, an element defined by the statement "including a..." does not exclude the presence of additional identical elements in the process, method, article or device including such element. In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0134] As described above, it is only the implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for monitoring the safety distance in power distribution construction, characterized in that, The method includes: Obtaining a power distribution operation scene image of a power distribution operation site by using a binocular vision image acquisition device; Inputting the power distribution operation scene image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation, and outputting a three-dimensional information disparity map; based on the binocular camera calibration parameters, converting the three-dimensional information disparity map into a power distribution operation depth map; Based on the power distribution operation scene image and a pre-constructed target detection neural network model, determining key operation elements and two-dimensional coordinate information; at least the key operation elements include operation personnel, construction machinery, and energized equipment; Based on the power distribution operation depth map, the two-dimensional coordinate information, and preset depth threshold information, screening the depth values of the pixel points within the target frame of the key operation elements to obtain the image coordinates of the screened pixel points; Converting the image coordinates of the screened pixel points and the depth values into a coordinate system to obtain the initial three-dimensional point cloud data of the key operation elements; performing clustering screening on the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; Using a preset K-D tree algorithm to determine the first nearest point coordinates between the operator point cloud and the construction machinery point cloud, the second nearest point coordinates between the operator point cloud and the energized equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the energized equipment point cloud from the effective three-dimensional point cloud data; Based on the first nearest point coordinates, the second nearest point coordinates, and the third nearest point coordinates, determining the safety distances between the operation personnel, the construction machinery, and the energized equipment.

2. The method according to claim 1, wherein The pre-constructed stereo matching neural network model is the IGEV++ model; The step of inputting the power distribution operation scene image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation and outputting a three-dimensional information disparity map includes: Adopting a multi-range strategy, calculating the feature correlation volume group by group in a small disparity range, and sparsely calculating the feature correlation volume through adaptive patch matching in a large disparity range to determine the matching degree at different disparities; Using a lightweight 3D convolutional network model to regularize the multi-range cost volume, and fusing multi-scale geometric context information to obtain an aggregated cost volume; Performing weighted summation on the aggregated cost volume through a Soft-argmin operation, calculating continuous disparity values, and generating an initial disparity map; Based on ConvGRU for iterative update, combining a selective feature fusion module, and using multi-range geometric features to optimize and correct the initial disparity map to obtain the three-dimensional information disparity map.

3. The method according to claim 1, wherein The step of determining key operation elements and two-dimensional coordinate information based on the power distribution operation scene image and a pre-constructed target detection neural network model includes: Inputting the power distribution operation scene image into a pre-constructed target detection neural network model for recognition and analysis to locate the key operation elements; Outputting the target frame of the key operation elements and the two-dimensional coordinate information thereof in the image coordinate system; The step of converting the three-dimensional information disparity map into a power distribution operation depth map based on the binocular camera calibration parameters includes: Using the binocular camera calibration parameters, convert the disparity value in the three-dimensional information disparity map into the power distribution operation depth map including the depth values of each point in the power distribution operation site; wherein, the conversion formula is: ； In the formula, is the depth value of each point in the depth map of the power distribution operation; is the focal length of the binocular vision image acquisition device; is the horizontal baseline distance of the binocular vision image acquisition device; is the disparity value.

4. The method according to claim 1, wherein The pre-constructed target detection neural network model is the D-FINE model. The network uses HGNetv2 as the feature extraction backbone structure, and the encoder-decoder of the Transformer architecture is connected to process features of different scales. Finally, the model prediction results are output through the traditional regression head and the D-FINE head.

5. The method according to claim 1, characterized in that The coordinate system conversion formula is: ； In the formula, and are the focal lengths of the camera in the , directions respectively; ([[]] , ) are the image coordinates of the filtered pixel points; ([[]] , ) are the central coordinates of the imaging plane of the binocular camera; R and T are the external parameter rotation matrix and offset matrix of the camera; ([[]] , , ) are the initial three-dimensional point cloud data.​​​ 6. The method according to claim 1, wherein The clustering and screening algorithm is the Mean Shift algorithm. By iteratively calculating the density-weighted mean within the neighborhood of each point, the points are gradually moved to the local density maximum points. Finally, the points converging to the same peak are grouped into one cluster, and the cluster structure can be automatically identified.

7. The method according to claim 1, characterized in that, The coordinates of the first nearest point are , ; The coordinates of the second nearest point are , ; The coordinates of the third nearest point are , ; The safety distance calculation formula is: ； ； ； Wherein, d1 represents the Euclidean distance between the point cloud of the operator and the point cloud of the construction machinery, d2 represents the Euclidean distance between the point cloud of the operator and the point cloud of the energized equipment, and d3 represents the Euclidean distance between the point cloud of the construction machinery and the point cloud of the energized equipment.

8. A distribution construction safety distance monitoring device, characterized in that, The device includes: An acquisition module, configured to acquire the power distribution operation scene image of the power distribution operation site by using a binocular vision image acquisition device; A calculation module, configured to input the power distribution operation scene image into a pre-constructed stereo matching neural network model for binocular stereo matching calculation, and output a three-dimensional information disparity map; based on the binocular camera calibration parameters, convert the three-dimensional information disparity map into a power distribution operation depth map; A determination module, configured to determine key operation elements and two-dimensional coordinate information based on the power distribution operation scene image and a pre-constructed target detection neural network model; at least the key operation elements include operators, construction machinery, and energized equipment; A screening module, configured to screen the depth values of the pixel points within the target frame of the key operation elements based on the power distribution operation depth map, the two-dimensional coordinate information, and preset depth threshold information, and obtain the image coordinates of the screened pixel points; A conversion module, configured to perform coordinate system conversion on the image coordinates of the screened pixel points and the depth values to obtain the initial three-dimensional point cloud data of the key operation elements; perform clustering and screening on the initial three-dimensional point cloud data to obtain effective three-dimensional point cloud data; The determination module is further configured to use a preset K-D tree algorithm to determine the first nearest point coordinates between the operator point cloud and the construction machinery point cloud, the second nearest point coordinates between the operator point cloud and the energized equipment point cloud, and the third nearest point coordinates between the construction machinery point cloud and the energized equipment point cloud from the effective three-dimensional point cloud data; The determination module is further configured to determine the safety distance between the operator, the construction machinery, and the energized equipment based on the first nearest point coordinates, the second nearest point coordinates, and the third nearest point coordinates.

9. A power distribution construction safety distance monitoring device, characterized in that, Includes: A memory, configured to store executable instructions; A processor, configured to implement the power distribution construction safety distance monitoring method according to any one of claims 1 to 7 when executing the executable instructions stored in the memory.

10. A computer-readable storage medium stores executable instructions for causing a processor to implement the power distribution construction safety distance monitoring method according to any one of claims 1 to 7 when executing the executable instructions.

Images