Unmanned aerial vehicle inspection method for power distribution network and electronic equipment

Abstract

The invention provides an unmanned aerial vehicle inspection method for a power distribution network and electronic equipment. The method comprises the following steps: acquiring a three-dimensional model and an inspection task of the power distribution network; wherein the three-dimensional model is pre-constructed according to satellite remote sensing data and unmanned aerial vehicle remote sensing data of the power distribution network; based on the inspection task, utilizing the three-dimensional model to determine an inspection path of the inspection unmanned aerial vehicle; and performing routing inspection by using the routing inspection unmanned aerial vehicle according to the routing inspection path to obtain a routing inspection result. According to the unmanned aerial vehicle inspection method for the power distribution network and the electronic equipment provided by the invention, autonomous planning and autonomous inspection of the unmanned aerial vehicle inspection path of the power distribution network are realized, and the automation degree and efficiency of unmanned aerial vehicle inspection are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of drone inspection technology, and in particular to a drone inspection method and electronic equipment for power distribution networks. Background Art

[0002] Traditional power grid inspections require significant manpower and material resources and are inefficient. However, the use of drones can significantly improve the efficiency and timeliness of grid inspections. Grid-based drone deployment enables comprehensive, frequent inspections of the power grid, enabling timely identification and resolution of issues, thus saving grid companies significant costs. Therefore, drone inspections have a wide range of applications in distribution networks. However, due to the wide coverage, complex grid structure, and shortage of human resources for operations and maintenance, the application of drone systems in distribution networks still faces numerous limitations, including insufficient automation and low inspection efficiency. Summary of the Invention

[0003] In view of this, the purpose of this application is to propose a drone inspection method and electronic equipment for distribution networks to solve the problems of insufficient automation and low inspection efficiency in the existing technology.

[0004] Based on the above objectives, the present application provides a method for drone inspection of a power distribution network, comprising:

[0005] Obtaining a three-dimensional model of the power distribution network and inspection tasks; wherein the three-dimensional model is pre-constructed based on satellite remote sensing data and drone remote sensing data of the power distribution network;

[0006] Determine an inspection path for the inspection drone using the three-dimensional model based on the inspection task;

[0007] The inspection drone is used to perform inspections along the inspection path to obtain inspection results.

[0008] Furthermore, determining the inspection path of the inspection drone using the three-dimensional model based on the inspection task includes:

[0009] Determine the inspection target and the inspection drone based on the inspection task, and obtain the path planning constraints corresponding to the inspection drone;

[0010] Using the three-dimensional model to obtain the location information of each inspection target and the location information of the obstacle;

[0011] The inspection path is determined based on the position information of each inspection target, the position information of the obstacle, and the path planning constraint condition.

[0012] Further, the path planning constraint conditions include a maximum flight range constraint condition, a minimum step size constraint condition, and a maximum path deflection angle constraint condition;

[0013] Determining the inspection path based on the position information of each inspection target, the position information of the obstacle, and the path planning constraint conditions includes:

[0014] Obtaining the preset shooting distance of the inspection target;

[0015] Based on the position information of each inspection target, the preset shooting distance, and the position information of the obstacle, determining at least one shooting position information corresponding to each inspection target;

[0016] Under the constraints of the maximum flight range constraint condition, the minimum step size constraint condition, and the maximum path deflection angle constraint condition, generating the inspection path by using a shortest path planning optimization algorithm integrated with an obstacle avoidance technology based on each shooting position information and the position information of the obstacle.

[0017] Further, using the inspection drone to perform inspections according to the inspection path to obtain inspection results includes:

[0018] Using the inspection drone to collect real-time images of the inspection target according to the inspection path to obtain inspection images;

[0019] Using an object detection algorithm and an image classification algorithm to identify the defect types of the inspection target in the inspection images to obtain inspection results.

[0020] Further, using the object detection algorithm and the image classification algorithm to identify the defect types of the inspection target in the inspection images to obtain inspection results includes:

[0021] Using the object detection algorithm to determine the target boundary of the inspection target in the inspection image, and cropping the inspection image based on the target boundary to obtain a target image;

[0022] Using the image classification algorithm to classify the defect types of the target image to obtain inspection results.

[0023] Further, using the object detection algorithm to determine the target boundary of the inspection target in the inspection image includes:

[0024] Using a multi-scale feature extraction network to extract multi-level features of the inspection image;

[0025] Using a self-supervised attention mechanism to enhance the features of the multi-level features of the inspection image to obtain enhanced features;

[0026] Fuse the enhanced features with the multi-level features of the inspection image to obtain fused features;

[0027] Generate candidate bounding boxes based on the fused features using a region proposal network;

[0028] Output the target boundary after performing regression correction and non-maximum suppression operations on the candidate bounding boxes.

[0029] Further, the defect type classification of the target image using the image classification algorithm to obtain the inspection result includes:

[0030] Extract the depth features of the target image using a feature extraction network;

[0031] Input the extracted depth features of the target image into a fully connected layer for defect type classification, and output the defect type corresponding to the target image.

[0032] Further, the three-dimensional model is constructed through the following steps:

[0033] Obtain the satellite remote sensing data of the power distribution network taken by a low-earth orbit satellite;

[0034] Use a drone to take an oblique photograph of the power distribution network to obtain the drone remote sensing data;

[0035] Use image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data to obtain fused image data;

[0036] Generate the three-dimensional model of the power distribution network based on the fused image data using three-dimensional reconstruction technology.

[0037] Further, before using image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data, it further includes:

[0038] Preprocess the satellite remote sensing data, and the preprocessing includes one or more of radiometric calibration, geometric correction, and atmospheric correction.

[0039] Based on the same inventive concept, the present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the above-mentioned method for drone inspection of a power distribution network.

[0040] As can be seen from the above, a method and an electronic device for unmanned aerial vehicle (UAV) inspection of a power distribution network provided by this application pre-construct a three-dimensional model of the power distribution network through satellite remote sensing data and UAV remote sensing data, effectively integrating the large-scale environmental information contained in the satellite remote sensing data and the specific information of the power distribution equipment of the power distribution network in the UAV remote sensing data. Compared with modeling using data from a single source, the constructed three-dimensional model can contain richer information and be more accurate, thereby providing more accurate and precise information for the planning of the UAV inspection path; because the three-dimensional model of the power distribution network contains rich and accurate information, the obtained inspection path is also more accurate. The inspection UAV inspects the power distribution network according to the inspection path, can effectively complete the inspection task, realizes the autonomous planning and autonomous inspection of the UAV inspection path, and effectively improves the inspection efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in this application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only the embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0042] Figure 1 Schematic diagram of a method for UAV inspection of a power distribution network according to an embodiment of this application;

[0043] Figure 2 Schematic diagram of the UAV flight path according to an embodiment of this application;

[0044] Figure 3 Schematic diagram of a device for UAV inspection of a power distribution network according to an embodiment of this application;

[0045] Figure 4 Schematic diagram of an electronic device according to an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following further elaborates on this application in detail with reference to specific embodiments and the accompanying drawings.

[0047] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of this application should have the ordinary meanings understood by those of ordinary skill in the field to which this application belongs. The "first", "second" and similar terms used in the embodiments of this application do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0048] Grid inspection is the regular or real-time inspection of facilities such as transmission lines, substations, and distribution equipment in the power system, aiming to detect problems such as equipment aging, damage, foreign object interference, and environmental hazards to ensure the safe and stable operation of the grid. Traditional inspections usually rely on manual climbing, ground patrols, or helicopter inspections, which require a large amount of manpower and material resources and have problems such as low efficiency, high risk, and high cost. Compared with traditional manual inspection methods, unmanned aerial vehicle (UAV) power inspection has the advantages of being faster, more convenient, and safer, which can greatly improve the efficiency and timeliness of grid inspections. Through the grid-based deployment of UAV automatic airports, it is possible to achieve all-round and high-frequency inspections of the grid, timely discover and handle problems, thus saving a large amount of costs for grid enterprises. Therefore, UAV inspections have a wide range of application scenarios in the distribution network.

[0049] However, due to the characteristics of wide distribution network coverage, complex grid structure, and shortage of operation and maintenance human resources, there are still many limitations in the application of UAV systems in the distribution network, such as:

[0050] (1) The efficiency and accuracy of 3D map modeling are not high

[0051] Existing distribution network modeling is usually constructed only using image data from a single source. Due to the limitations of acquisition means, the accuracy and precision of the modeling are limited, and the modeling efficiency is also limited, which cannot provide accurate data for subsequent UAV inspections and affects the inspection efficiency of UAVs.

[0052] (2) The recognition accuracy of detection targets is limited

[0053] There are many types of targets that need to be inspected on distribution networks, such as insulators, wire clamps, and bobbins. These targets are numerous and widely distributed within distribution networks. For example, insulators are widely distributed between conductors, crossarms, and towers, providing excellent insulation and forming a crucial component of transmission lines. However, due to long-term exposure to the natural environment, strong electric fields, and excessive mechanical loads, insulators are prone to failures such as string drop, breakage, lightning strikes, and foreign objects. Once a problem occurs, it can seriously impact the stable operation of the power system. However, due to the numerous types, large number, and widespread distribution of insulators, existing drone inspections struggle to accurately capture and identify insulators amidst numerous obstacles, impacting their effectiveness.

[0054] (3) Unable to identify the defect type of the detection target

[0055] Existing drone inspections primarily rely on capturing images of the target object, but are unable to identify the object's condition or the type of defect. Drones capture a large amount of image and video footage during inspections, leaving power workers to visually determine whether the target is functioning properly, whether any faults have occurred, and what the fault type is. This is inefficient, especially when dealing with large amounts of data. It is also prone to errors, preventing timely detection of grid issues and potentially creating safety risks.

[0056] These issues are even more pronounced in renewable energy distribution networks. These networks, which include distributed energy sources like wind and photovoltaic power, are often characterized by wide distribution areas, remote locations, and complex environments, such as mountainous areas. Furthermore, the diversity and widespread distribution of renewable energy equipment pose additional limitations and challenges to 3D map construction and target recognition.

[0057] In view of this, the present application provides a UAV inspection method for power distribution network, which can be applied to new energy power distribution network, realizes autonomous path planning of UAV, and can effectively improve efficiency, such as Figure 1 As shown, the method includes:

[0058] S101. Obtain a three-dimensional model of a power distribution network and inspection tasks; wherein the three-dimensional model is pre-constructed based on satellite remote sensing data and drone remote sensing data of the power distribution network;

[0059] Specifically, satellite remote sensing data is collected by low-earth orbit satellites. The image data collected by satellites has a wide coverage area and is suitable for large-area monitoring. UAV remote sensing data is collected by UAVs equipped with image data collection devices. The UAV remote sensing data collected by UAVs has a high resolution and can effectively capture details of power distribution equipment (such as insulators, wire clips, spinning hammers, etc.) in the power distribution network, can effectively identify the power distribution equipment, and obtain its location information. The new energy power distribution network has a wide distribution range and complex terrain. Using low-earth orbit satellites can collect more comprehensive overall data of the new energy power distribution network and is not prone to omissions. The areas where the power distribution network is located are usually remote and sparsely populated, which is conducive to the growth of vegetation. Especially for the new energy power distribution network, the growth of vegetation will not only affect the normal operation of the transmission line, but also increase the difficulty of subsequent inspections, and will even affect the inspection path when necessary. The data collected by low-earth orbit satellites can be revisited regularly, that is, satellite remote sensing data with time-series data can be provided. By obtaining the satellite remote sensing data with time-series data collected by low-earth orbit satellites regularly or irregularly in real time, the environmental changes of the power distribution network, such as the growth of vegetation, can be obtained in a timely manner. Thus, the three-dimensional model can be updated regularly, thereby improving the accuracy and precision of the three-dimensional model. The three-dimensional model of the power distribution network constructed from satellite remote sensing data and UAV remote sensing data can include the environmental data of the power distribution network and the detailed data of the power distribution equipment, providing a basis for the accurate planning of the subsequent UAV inspection path.

[0060] S102. Determine the inspection path of the inspection UAV based on the inspection task using the three-dimensional model;

[0061] Specifically, the inspection tasks include daily inspections, fault inspections, and special power protection inspections. According to the different inspection tasks, corresponding inspection objectives and inspection cycles can be pre-configured. According to different actual needs, the inspection tasks can be further refined. For example, according to the power distribution equipment, exemplary inspections include insulator inspections, tower inspections, line inspections, external hidden danger inspections, etc., without specific limitations. After determining the inspection task, the inspection path of the inspection UAV can be planned based on the three-dimensional model.

[0062] S103. Use the inspection UAV to perform inspections according to the inspection path to obtain inspection results.

[0063] Specifically, the inspection UAV is a UAV equipped with an image collection device. After obtaining the inspection path, the inspection UAV can be launched. The inspection UAV flies according to the inspection path to perform inspections, thereby obtaining inspection results and realizing the autonomous inspection of the power distribution network by the UAV.

[0064] In this application, through steps S101 to S103, a three-dimensional model of the power distribution network is pre-constructed using satellite remote sensing data and unmanned aerial vehicle (UAV) remote sensing data. The large-scale environmental information contained in the satellite remote sensing data and the specific information of the power distribution equipment of the power distribution network in the UAV remote sensing data are effectively integrated. Compared with modeling using data from a single source, the constructed three-dimensional model can contain richer information and be more accurate. Therefore, it can provide more accurate and precise information for the planning of the UAV inspection path. Since the three-dimensional model of the power distribution network contains rich and accurate information, the inspection path obtained using it is also more accurate. The inspection UAV conducts inspections on the power distribution network according to the inspection path, can effectively complete the inspection task, realize the autonomous planning and autonomous inspection of the UAV inspection path, and effectively improve the inspection efficiency and accuracy.

[0065] In some embodiments, determining the inspection path of the inspection UAV using the three-dimensional model based on the inspection task includes:

[0066] S1021. Determine the inspection target and the inspection UAV based on the inspection task, and obtain the path planning constraint conditions corresponding to the inspection UAV;

[0067] Specifically, according to different inspection tasks, inspection targets and suitable inspection UAVs are preset. The inspection targets include power distribution equipment on the power distribution network, such as insulators, wire clamps, spinning hammers, lines, and poles. Corresponding inspection targets can be preset according to different tasks. Power inspection UAVs mainly use rotor UAVs and fixed-wing UAVs. The main difference between the two lies in whether they can hover and take off and land in place. According to the specific characteristics of rotor UAVs and fixed-wing UAVs, the inspection tasks suitable for them can be preset. For example, fixed-wing UAVs are suitable for large-scale line corridor inspection tasks, and multi-rotor UAVs are suitable for precise pole inspection tasks. When the UAV constructs the inspection path, it needs to be constrained by its path planning constraint conditions, and there are certain differences in the path planning constraint conditions of different models and characteristics of inspection UAVs. Therefore, after determining the inspection UAV for inspection, the corresponding path planning constraint conditions are further obtained, so that the subsequent generated inspection path can conform to the flight and performance characteristics of the inspection UAV, ensuring the smooth realization of autonomous inspection.

[0068] S1022. Use the three-dimensional model to obtain the position information of each inspection target and the position information of obstacles;

[0069] Specifically, in the constructed 3D model, it includes the distribution equipment information and environmental composition information of the distribution network. The environmental composition information can include information such as trees and buildings. During the inspection process of the distribution network, objects such as trees and buildings are the targets that the inspection UAV needs to bypass. Therefore, objects that affect the flight of the inspection UAV can be set as obstacles. For example, they can be classified according to the type of object. All trees, buildings, etc. can be classified as obstacles, or all objects other than the inspection targets can be regarded as obstacles, without specific restrictions. In the 3D model, the distribution equipment information includes the equipment type and its location information, and the environmental composition information also includes the object type and its location information. Therefore, after determining the inspection targets, the location information of the inspection targets and the location information of the obstacles can be obtained according to the 3D model, providing detailed and reliable location information for the subsequent planning of the inspection path.

[0070] S1023. Determine the inspection path based on the location information of each inspection target, the location information of the obstacles, and the path planning constraint conditions.

[0071] Specifically, after obtaining the location information of each inspection target and the location information of the obstacles, a combined inspection path can be generated with the path planning constraint conditions.

[0072] In this application, according to steps S1021 - S1023, since the 3D model provides the type information and location information of the distribution equipment constituting the distribution network and the location information of the obstacles, and each inspection task has a preset inspection target, the inspection target can be automatically determined after determining the inspection task. Thus, the location information of each inspection target and the location information of the obstacles can be automatically obtained using the 3D model, determining the locations where the inspection UAV needs to conduct inspections and the locations where it needs to bypass. Then, combined with the path planning constraint conditions that conform to the performance characteristics of the inspection UAV itself, an inspection path suitable for the inspection UAV can be obtained, realizing the autonomous planning of the inspection path and providing a prerequisite for the autonomous inspection of the UAV.

[0073] In some embodiments, the path planning constraint conditions include a maximum flight range constraint condition, a minimum step size constraint condition, and a maximum path deflection angle constraint condition;

[0074] The determining of the inspection path based on the location information of each inspection target, the location information of the obstacles, and the path planning constraint conditions includes:

[0075] Obtain the preset shooting distance of the inspection target;

[0076] Based on the location information of each inspection target, the preset shooting distance, and the location information of the obstacles, determine at least one shooting location information corresponding to each inspection target;

[0077] Under the constraints of the maximum flight range constraint condition, the minimum step size constraint condition, and the maximum path deflection angle constraint condition, based on each shooting position information and the position information of the obstacle, the shortest path planning optimization algorithm integrated with obstacle avoidance technology is used to generate the inspection path.

[0078] The maximum flight range constraint condition refers to the farthest distance that a UAV can cover in a single flight, which is usually determined by factors such as battery capacity and energy efficiency; the minimum step size constraint condition refers to the minimum allowable distance between adjacent waypoints in path planning, which is usually determined by the control accuracy of the UAV, the resolution of the navigation system, or task requirements; the maximum path deflection angle constraint condition refers to the maximum steering angle allowed between adjacent flight segments of the UAV, which is limited by the maneuverability of the UAV (such as turning radius, angular velocity) and safety requirements.

[0079] (1) Maximum flight range constraint condition

[0080] During the execution of the inspection task, the UAV is restricted by battery life and has a maximum flight range, denoted as V max , as Figure 2 shown, assuming that a certain flight path has a total of n nodes, and the flight range of the i-th segment of the path is denoted as V i , then the total flight range V of this path must satisfy the following formula:

[0081]

[0082] (2) Minimum step size constraint condition

[0083] When the UAV changes its current flight attitude, it needs to fly a certain distance to overcome the influence of inertia, and the minimum value of this distance is called the minimum step size, denoted as L min . The distance l j between any two waypoints must satisfy the size constraint:

[0084] l j ≥L min

[0085] (3) Maximum path deflection angle constraint condition

[0086] The azimuth deflection angle of the current flight segment relative to the previous flight segment is called the path deflection angle. Among them, r min is the minimum turning radius, and L min is the minimum flight path step size. Due to the limitation of the UAV's maneuverability, the path deflection angle Δφ i of the i-th segment must satisfy:

[0087]

[0088] Among them, φ max is the maximum path deflection angle:

[0089]

[0090] For different types of drones, there are certain differences in their battery capacity, operating characteristics, and maneuverability. The specific values of the maximum flight range constraint condition, minimum step size constraint condition, and maximum path deflection angle constraint condition are also different. Specifically, the corresponding maximum flight range constraint condition, minimum step size constraint condition, and maximum path deflection angle constraint condition can be preset according to the type of inspection drone.

[0091] For different inspection targets, the required shortest shooting distance may be different. For example, for some inspection targets with small volumes, a closer shooting distance is required. For some inspection targets with large volumes and easy to identify, a farther shooting distance can be set to reduce the flight path of the inspection drone and lower the flight difficulty of the inspection drone. Therefore, the corresponding preset shooting distance can be set according to different inspection targets. After obtaining the position information of each inspection target, at least one shooting position information corresponding to each inspection target can be determined by combining its corresponding preset shooting distance and the position information of the obstacle. The shooting position identified by the shooting position information is the position where the inspection drone collects the image of the inspection target. Specifically, an inspection target can correspond to one shooting position information or multiple shooting position information, and there is no specific limitation. Exemplarily, an inspection target can correspond to four shooting position information, that is, four shooting position information are determined by surrounding the inspection target once in the horizontal direction, so that images in four directions of the inspection target can be obtained, avoiding the problem of unclear observation caused by a single angle or a single image, and thus realizing more efficient inspection.

[0092] In the path planning process of the drone, it is first necessary to determine multiple waypoints, and then set the drone to drive straight automatically between the waypoints, and then the flight path of the drone is obtained. In this application, the shooting position information corresponding to each inspection target is the waypoint of the inspection drone, and the position corresponding to the position information of the obstacle is the obstacle point that the drone needs to bypass. Under the constraints of the maximum flight range constraint condition, minimum step size constraint condition, and maximum path deflection angle constraint condition, combined with the waypoints to be passed and the obstacle points to be bypassed, the final inspection path can be generated by using the shortest path planning optimization algorithm integrated with obstacle avoidance technology.

[0093] The shortest path planning optimization algorithm integrated with obstacle avoidance technology is an intelligent decision-making method that combines path length minimization and dynamic or static obstacle avoidance. Specifically, the execution steps of the shortest path planning optimization algorithm integrated with obstacle avoidance technology include:

[0094] Based on each shooting position information, use the shortest path planning algorithm (such as the improved ant colony algorithm, A-Star algorithm, or rapidly-exploring random tree algorithm, etc.) to generate the shortest connected path. Transform the maximum flight range constraint condition, minimum step size constraint condition, and maximum path deflection angle constraint condition into mathematical constraints through a cost function, and then insert obstacle avoidance waypoints or adjust the path segments according to the position information of the obstacles to avoid the safe area of the obstacles and generate the final inspection path.

[0095] During the planning process of the inspection path, safety constraint conditions and efficiency constraint conditions can be further incorporated.

[0096] (1) Safety constraint conditions

[0097] The safety constraint conditions of the UAV inspection path include two aspects: the UAV inspection line is severely affected by the electromagnetic interference around it, and the gyroscope and GPS will be greatly affected, so the safety of the UAV flight itself must be ensured; the UAV must maintain a distance from the line and the tower during the inspection to ensure the safety of the inspection object.

[0098] (2) Efficiency constraint conditions

[0099] When the UAV goes out for inspection, it will be subject to various restrictive factors, so its inspection efficiency must be ensured. Under the premise of meeting the requirements of the inspection task, avoid the repetition of the inspection path; when planning each waypoint, make the distance between waypoints close, shorten the length of the inspection path, and improve the inspection efficiency; on the premise of meeting safety, increase the flight speed of the UAV and shorten the inspection duration.

[0100] In some embodiments, using the inspection UAV to perform inspection according to the inspection path to obtain an inspection result, including:

[0101] Using the inspection UAV to collect real-time images of the inspection target according to the inspection path to obtain inspection images;

[0102] Using the target detection algorithm and image classification algorithm to identify the defect types of the inspection target in the inspection images to obtain the inspection result.

[0103] In this application, a patrol drone is used to collect real-time images of the patrol target according to the patrol path. When there are faults or defects in the patrol target, the corresponding fault and defect information can be captured in the real-time images in a timely manner. Then, the target detection algorithm and the image classification algorithm are used to identify the defect type of the patrol target, so as to obtain the patrol result. Specifically, when the patrol target is normal, the recognition result of the corresponding defect type can be classified as normal. When there are notches, fractures, rust, etc. on the surface of the patrol target, the corresponding defect types are notches, fractures, rust, etc. Thus, the automatic recognition of the defect type of the patrol target is realized, without manual judgment, improving the degree of patrol automation and patrol efficiency, and reducing the labor cost.

[0104] In some embodiments, the using of the target detection algorithm and the image classification algorithm to identify the defect type of the patrol target in the patrol image and obtain the patrol result includes:

[0105] Using the target detection algorithm to determine the target boundary of the patrol target in the patrol image, and cropping the patrol image based on the target boundary to obtain a target image;

[0106] Using the image classification algorithm to classify the defect type of the target image to obtain the patrol result.

[0107] In this application, first use the target detection algorithm to determine the position of the patrol target in the patrol image, and then use the image classification algorithm to identify the defect type, so that when performing image classification, it can focus on the patrol target, with higher classification accuracy and more accurate patrol results. Specifically, after obtaining the patrol image, the target detection algorithm can identify the object classification and the corresponding coordinates included in the patrol image, and finally output the coordinates corresponding to the object classification as the patrol target as the target boundary, and then crop the target image according to the target boundary, so as to remove the information irrelevant to the patrol target in the patrol image; when using the target image to classify the defect type, since the information irrelevant to the patrol target has been removed, the classification interference can be effectively reduced, thereby improving the classification accuracy.

[0108] Object detection algorithms can be further divided into two-step object detection algorithms and one-step object detection algorithms. The two-step object detection algorithm extracts the region of interest (ROI) containing the inspection target from the inspection image through a region extraction structure, then classifies and locates the ROI containing the inspection target, and finally outputs the target ratio boundary. Due to the screening of the region position, the detection effect of the two-step object detection algorithm is often more accurate, but the screening of the region position will consume a certain amount of time. Therefore, the two-step object detection algorithm requires a longer inference time. The one-step object detection algorithm directly uses the regression method to predict the coordinate information and class information without region selection, that is, it can output the target boundary with one operation, so it has a faster detection speed but lower detection accuracy. Specifically, the one-step object detection algorithm or the two-step object detection algorithm can be selected according to actual needs, application scenarios, etc., and there is no specific limitation.

[0109] In some embodiments, using the object detection algorithm to determine the target boundary of the inspection target in the inspection image includes:

[0110] Extracting multi-level features of the inspection image using a multi-scale feature extraction network;

[0111] Using a self-supervised attention mechanism to enhance the multi-level features of the inspection image to obtain enhanced features;

[0112] Fusing the enhanced features with the multi-level features of the inspection image to obtain fused features;

[0113] Generating candidate bounding boxes based on the fused features using a region generation network;

[0114] Performing regression correction and non-maximum suppression operations on the candidate bounding boxes and then outputting the target boundary.

[0115] The multi-scale feature extraction network is obtained by using a pre-trained ResNet as the basic network and then adding multiple convolutional layers with different scales. The following takes adding convolutional layers of 3x3, 5x5, and 7x7 as an example to further illustrate the specific calculation process of using the multi-scale feature extraction network to extract multi-level features of the inspection image:

[0116] First, input the inspection image into ResNet to extract the feature map F res ;

[0117] Then input F res into multiple convolutional layers with different scales (3x3, 5x5, 7x7) for feature extraction. The expressions of the multiple convolutional layers with different scales are: F k =Conv k (F res ), k∈{3, 5, 7}; where Convk Denotes a convolution operation using a k×k convolution kernel;

[0118] Finally, the features extracted from multiple convolutional layers of different scales are concatenated to obtain multi-level feature F ms , that is: F ms = Concat(F3, F5, F7).

[0119] The self-supervised attention mechanism is a technique that combines self-supervised learning and the attention mechanism. It learns the internal structure of data in a self-supervised manner and uses the attention mechanism to dynamically allocate weights, thereby improving the ability to understand and process data. Self-supervised learning is an unsupervised learning method that constructs pseudo-labels to guide the model to learn. This method does not require external labeled data but uses the structure and characteristics of the data itself to generate learning objectives. The attention mechanism is a mechanism that simulates human attention. It allows the model to dynamically allocate weights when processing data, thereby focusing on more important parts. The self-supervised attention mechanism combines the advantages of self-supervised learning and the attention mechanism, generates pseudo-labels through self-supervised tasks, and uses the attention mechanism to dynamically allocate weights, thereby improving the ability to understand and process data. By introducing the self-supervised attention mechanism to process the multi-level features of the inspection images, more attention can be paid to the parts related to the inspection target, making the weights of the parts related to the inspection target in the fused features higher, thus providing a prerequisite for the accurate generation of subsequent candidate bounding boxes.

[0120] After fusing the multi-level features with the enhanced features, the fused features are obtained, and then candidate bounding boxes are generated based on the fused features using the region proposal network. The region proposal network (RPN) automatically generates candidate regions that may contain the target through deep learning, that is, candidate bounding boxes containing the inspection target. The candidate boundaries generated by the region proposal network are usually not accurate enough and need to be refined through regression and eliminate overlapping redundant boxes through non-maximum suppression operations, thereby obtaining the position where the target boundary can accurately cover the inspection target, providing a basis for the accurate classification of subsequent defect types.

[0121] In some embodiments, the using the image classification algorithm to classify the defect type of the target image to obtain the inspection result includes:

[0122] Using a feature extraction network to extract the deep features of the target image;

[0123] Inputting the extracted deep features of the target image into a fully connected layer for defect type classification, and outputting the defect type corresponding to the target image.

[0124] Specifically, the image classification algorithm is implemented through a feature extraction network and a fully connected layer. The deep features of the target image are extracted through the feature extraction network, and the fully connected layer maps the deep features obtained by the feature extraction network to a vector representation of a fixed length. The number of neurons in the fully connected layer is equal to the number of categories to be classified in the data. The softmax function is used at the end of the fully connected layer to obtain the probability results of each category. Finally, the category corresponding to the highest category probability is output as the defect type corresponding to the target image, thereby realizing the automatic identification of the defect type. Through the image classification algorithm, the accurate identification of the inspection target defects is realized, eliminating the need for manual identification, improving the inspection efficiency and inspection automation, and reducing the labor cost.

[0125] In some embodiments, the three-dimensional model is constructed through the following steps:

[0126] Obtain the satellite remote sensing data of the power distribution network taken by the low-earth orbit satellite;

[0127] Use a drone to take an oblique photograph of the power distribution network to obtain the drone remote sensing data;

[0128] Use image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data to obtain fused image data;

[0129] Generate the three-dimensional model of the power distribution network based on the fused image data using three-dimensional reconstruction technology.

[0130] Specifically, when planning the flight path for the drone to take an oblique photograph of the power distribution network, the image heading overlap is 56% - 75%, and the side overlap is 30% - 35%. When the drone takes an oblique photograph of the power distribution network, it synchronously records information such as the longitude and latitude coordinates, flight speed, altitude, and direction angle of the image. Since the image format is small and the number of images obtained by a single aerial survey is considerable, the final drone remote sensing data is obtained by stitching and synthesizing the data from multiple aerial surveys.

[0131] After obtaining satellite remote sensing data and unmanned aerial vehicle (UAV) remote sensing data, image fusion technology is used to fuse them to obtain fused image data. Image fusion technology is a technology that processes multiple images from different sensors, different perspectives, or different time points to generate images with richer information and higher quality. Classified by fusion level, it includes pixel-level fusion, feature-level fusion, and decision-level fusion. Pixel-level fusion refers to directly operating on images at the pixel level, which can retain more detailed information such as edges and textures, and can be achieved through algorithms such as weighted average method, maximum value selection method, and wavelet transform method. Feature-level fusion requires first extracting the key features of the images (such as edges, shapes, textures, etc.) and then fusing these features. This method can reduce data redundancy and improve fusion efficiency, but may lose some details, and can be achieved through algorithms such as PCA (Principal Component Analysis) and SIFT (Scale-Invariant Feature Transform). Decision-level fusion refers to fusing at the highest level and making comprehensive decisions based on the classification or recognition results of the images, and can be achieved through algorithms such as Bayesian decision-making and fuzzy logic. In practical applications, the appropriate image fusion technology can be selected according to specific situations to achieve the fusion of satellite remote sensing data and UAV remote sensing data, and there is no specific limitation.

[0132] Specifically, the use of image fusion technology to fuse the satellite remote sensing data and the UAV remote sensing data to obtain fused image data includes:

[0133] (1) Unify the satellite remote sensing data and the UAV remote sensing data into the same coordinate system to achieve image registration between different remote sensing image data, and obtain the satellite data and UAV data after spatial registration. Commonly used coordinate systems include image coordinate system (introduced to describe the projection and transmission relationship of objects from the camera coordinate system to the image coordinate system during the imaging process, which is convenient to further obtain the coordinates in the pixel coordinate system, with the unit of m), imaging plane coordinate system, camera coordinate system (a coordinate system established on the camera, defined to describe the position of objects from the perspective of the camera, and serving as an intermediate link between the world coordinate system and the image / pixel coordinate system, with the unit of m), and world (absolute) coordinate system (a coordinate system of the three-dimensional world defined by the user, introduced to describe the position of the target object in the real world, with the unit of m). Through coordinate transformation calculation, the coordinate systems of different source spatial data are unified.

[0134] (2) Perform image fusion on the satellite data and UAV data after spatial registration to obtain fused image data with more abundant information, meeting the modeling requirements of the three-dimensional model of the power distribution network.

[0135] UAV remote sensing data includes point cloud data and panoramic image data. The initial image data obtained by the UAV's oblique photography is automatically processed by aerial triangulation to generate a large amount of point cloud data. The point cloud data is massive and contains all the information content in the three-dimensional scene, but it cannot provide physical information such as the material, texture, structure, and color of the object surface. The panoramic image data contains rich detailed features of the target surface. Therefore, the panoramic image data and the point cloud data can be used for joint registration to produce rich true three-dimensional products. After obtaining the fused image data, a three-dimensional model of the power distribution network is generated using three-dimensional reconstruction technology. The three-dimensional model contains the power distribution equipment information of the power distribution network, specifically including the equipment type and its location information.

[0136] In addition, in order to improve the accuracy and reliability of the three-dimensional model, laser scanning technology can also be used to conduct on-site measurements of the power distribution equipment of the power distribution network to obtain more accurate three-dimensional coordinate data. After completing the construction of the three-dimensional model, the actual measurement data can be used to verify and test it to check the accuracy and reliability of the model, so as to judge the applicability and accuracy of the model.

[0137] In some embodiments, before fusing the satellite remote sensing data and the UAV remote sensing data using the image fusion technology, it further includes:

[0138] Preprocessing the satellite remote sensing data, and the preprocessing includes one or several of radiometric calibration, geometric correction, and atmospheric correction.

[0139] Through radiometric calibration, the recorded original DN value (pixel brightness value) is converted into the surface reflectance outside the atmosphere, aiming to eliminate the error of the acquisition sensor itself and determine the accurate radiation value at the entrance of the acquisition sensor. The specific methods generally include laboratory calibration, on-board / on-satellite calibration, field calibration, etc. For different sensors, their radiometric calibration formulas are different. Through atmospheric correction, the radiance or surface reflectance is converted into the actual surface reflectance of the earth's surface to eliminate the errors caused by atmospheric scattering, absorption, and reflection.

[0140] Using a UAV equipped with a non-metric digital visible light device to conduct mapping of the power distribution network can achieve the acquisition of UAV remote sensing data. After the UAV uses the non-metric digital visible light device to collect the initial image data, it can output the UAV remote sensing data after operations such as preliminary processing, aerial triangulation, mapping image generation, and image accuracy analysis.

[0141] (1) Preliminary processing

[0142] Generally, the aerial survey of drones is equipped with non-metric digital visible light devices, and there are varying degrees of distortion in their lenses. This distortion causes the displacement of the coordinate points of the vectorized image data, which will affect the accuracy of aerial triangulation, thus causing deformation in the mapping of surveying and mapping, and preliminary processing is required; the preliminary processing includes one or several combinations of the camera calibration method, the bundle adjustment method, the polynomial transformation method, and the image registration method.

[0143] 1) Camera calibration method: The camera calibration method is to obtain the internal and external parameters of the camera by shooting the calibration plate image, and then perform geometric correction. Commonly used camera calibration methods include Zhang's calibration method, Chessboard calibration method, and circular calibration method.

[0144] 2) Bundle adjustment method: The bundle adjustment method is a geometric correction method based on the perspective projection model. By calculating the rays from the camera center to each pixel point on the image and using the corresponding relationship between these rays and the points in the real world to restore the geometric relationship of the image.

[0145] 3) Polynomial transformation method: The polynomial transformation method is a method of geometric correction of images using polynomial equations. By calculating the polynomial equations between the coordinates of the corresponding points in the image and using these equations to correct the geometric deformation of the image.

[0146] 4) Image registration method: The image registration method is a method of geometric correction by registering the drone image with a reference image. By finding the similarity of the corresponding points in the image and then using these corresponding points to calculate the transformation matrix between the images, and then registering the drone image to the reference image.

[0147] (2) Aerial triangulation

[0148] Using the ground control point data and image observation data to determine the orientation elements of the image, establish the corresponding strip or regional model, so as to obtain the encrypted points, plane coordinates, and elevations. Currently, the bundle adjustment method for regional network aerial triangulation is mainly adopted. Taking each image as a unit and the image point coordinates as the original observations, strictly based on the collinearity equation to generate the exterior orientation elements and ground coordinates, and after aerial triangulation, the mapping image of surveying and mapping can be generated.

[0149] (3) Mapping image generation

[0150] The mapping of surveying and mapping images includes DLG (Digital Line Graphic), DEM (Digital Elevation Model), and DOM (Digital Orthophoto Map). DLG is a geographic information vector dataset that expresses topographic elements in the form of points, lines, and surfaces or specific graphic symbols on a map, with the spatial relationships between various geographic elements and relevant attribute information. DEM digitally simulates the terrain through terrain elevation data and is a solid terrain model that represents ground elevation in the form of an ordered numerical array. DOM performs digital differential rectification and mosaicing on images and generates an orthophoto image set by cropping according to a certain map sheet range, having both the geometric accuracy of a map and the characteristics of an image.

[0151] (4) Image accuracy analysis

[0152] In the mapping of surveying and mapping images, the accuracy of DOM and DEM needs to be evaluated before use, and the evaluation results should meet relevant specifications.

[0153] The UAV inspection method and electronic device for a power distribution network provided in this application pre-construct a three-dimensional model of the power distribution network by integrating satellite remote sensing data and UAV remote sensing data, which can contain richer information about the environment and power distribution equipment, with more accurate modeling, providing a basis for the inspection path planning of subsequent inspection UAVs; determining the position information of each inspection target and the position information of obstacles using the three-dimensional model, and then further determining the position information of each shooting location, and then combining the constraints of the maximum flight range constraint condition, the minimum step size constraint condition, and the maximum path deflection angle constraint condition, using the shortest path planning optimization algorithm integrated with obstacle avoidance technology to generate the inspection path, realizing the autonomous planning of the inspection path; after obtaining the inspection path, using the inspection UAV to collect real-time images of the inspection targets according to the inspection path, and finally using the object detection algorithm and the image classification algorithm to identify the defect types of the inspection targets in the inspection images to obtain the inspection results, thus realizing the autonomous inspection of the UAV and the automatic identification of the defect types of the inspection targets, effectively improving the automation degree of UAV inspection, reducing manual operations, and improving the inspection efficiency.

[0154] It should be noted that the method of the embodiment of this application can be executed by a single device, such as a computer or a server. The method of this embodiment can also be applied to a distributed scenario and completed by the cooperation of multiple devices. In this case of a distributed scenario, one of the multiple devices can only execute one or more steps of the method of the embodiment of this application, and these multiple devices will interact with each other to complete the described method.

[0155] It should be noted that some embodiments of the present application have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the above embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0156] Based on the same inventive concept, corresponding to any of the above method embodiments, the present application further provides a drone inspection device for a power distribution network.

[0157] As Figure 3 shown, the device includes:

[0158] An acquisition module 301 for acquiring a three-dimensional model of the power distribution network and an inspection task; wherein, the three-dimensional model is pre-constructed according to satellite remote sensing data and drone remote sensing data of the power distribution network;

[0159] A path planning module 302 for determining an inspection path of an inspection drone based on the inspection task using the three-dimensional model;

[0160] An inspection module 303 for using the inspection drone to perform inspections according to the inspection path to obtain inspection results.

[0161] In some embodiments, the path planning module 302 is further configured to:

[0162] Determine an inspection target and the inspection drone based on the inspection task, and obtain path planning constraint conditions corresponding to the inspection drone;

[0163] Use the three-dimensional model to obtain the position information of each inspection target and the position information of obstacles;

[0164] Determine the inspection path based on the position information of each inspection target, the position information of the obstacles, and the path planning constraint conditions.

[0165] In some embodiments, the path planning module 302 is further configured to:

[0166] The determining the inspection path based on the position information of each inspection target, the position information of the obstacles, and the path planning constraint conditions includes:

[0167] Obtain a preset shooting distance of the inspection target;

[0168] Determine at least one shooting position information corresponding to each inspection target based on the position information of each inspection target, the preset shooting distance, and the position information of the obstacle;

[0169] Under the constraints of the maximum range constraint, the minimum step size constraint and the maximum path deflection angle constraint, the inspection path is generated based on each shooting position information and the position information of the obstacle using a shortest path planning optimization algorithm integrated with obstacle avoidance technology.

[0170] In some embodiments, the inspection module 303 is further configured to:

[0171] Using the inspection drone to collect real-time images of the inspection target along the inspection path to obtain an inspection image;

[0172] The target detection algorithm and the image classification algorithm are used to identify the defect type of the inspection target in the inspection image to obtain the inspection result.

[0173] In some embodiments, the inspection module 303 is further configured to:

[0174] Determining a target boundary of the inspection target in the inspection image by using the target detection algorithm, and cropping the inspection image based on the target boundary to obtain a target image;

[0175] The target image is classified into defect types using the image classification algorithm to obtain an inspection result.

[0176] In some embodiments, the inspection module 303 is further configured to:

[0177] Extracting multi-level features of the inspection image using a multi-scale feature extraction network;

[0178] Using a self-supervised attention mechanism to enhance the multi-level features of the inspection image to obtain enhanced features;

[0179] Fusing the enhanced features with the multi-level features of the inspection image to obtain fused features;

[0180] Generate a candidate bounding box using a region generation network based on the fused features;

[0181] The target boundary is output after regression correction and non-maximum suppression operations are performed on the candidate bounding box.

[0182] In some embodiments, the inspection module 303 is further configured to:

[0183] Extracting deep features of the target image using a feature extraction network;

[0184] Input the depth features of the obtained target image into the fully connected layer for defect type classification, and output the defect type corresponding to the target image.

[0185] In some embodiments, the apparatus further includes a model construction module for constructing the 3D model through the following steps:

[0186] Obtain the satellite remote sensing data of the power distribution network captured by the low-earth orbit satellite;

[0187] Use a drone to perform oblique photography on the power distribution network to obtain the drone remote sensing data;

[0188] Use image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data to obtain fused image data;

[0189] Generate the 3D model of the power distribution network based on the fused image data using 3D reconstruction technology.

[0190] In some embodiments, the model construction module further includes a preprocessing unit for preprocessing the satellite remote sensing data before fusing the satellite remote sensing data and the drone remote sensing data using image fusion technology, and the preprocessing includes one or more of radiometric calibration, geometric correction, and atmospheric correction.

[0191] For the convenience of description, when describing the above apparatus, it is divided into various modules according to functions for separate description. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0192] The apparatus in the above embodiment is used to implement the corresponding method for drone inspection of a power distribution network in any one of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated herein.

[0193] Based on the same inventive concept, corresponding to the method in any of the above embodiments, the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the method for drone inspection of a power distribution network described in any one of the above embodiments.

[0194] Figure 4 FIG. shows a more specific schematic diagram of the hardware structure of the electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. Among them, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other inside the device through the bus 1050.

[0195] The processor 1010 can be implemented in the form of a general - purpose CPU (Central Processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0196] The memory 1020 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 1020 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0197] The input / output interface 1030 is used to connect to the input / output module to implement information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Among them, the input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0198] The communication interface 1040 is used to connect to a communication module (not shown in the figure) to implement communication interaction between this device and other devices. Among them, the communication module can implement communication through a wired method (such as USB, network cable, etc.) or can implement communication through a wireless method (such as a mobile network, WIFI, Bluetooth, etc.).

[0199] The bus 1050 includes a path for transmitting information between various components of the device (such as the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0200] It should be noted that although the above - mentioned device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in the specific implementation process, this device may also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above - mentioned device may also only include the components necessary to implement the solutions of the embodiments of this specification and does not necessarily include all the components shown in the figure.

[0201] The electronic device of the above embodiment is used to implement the corresponding method for drone inspection of a power distribution network in any one of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated herein.

[0202] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the present application further provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute a method for drone inspection of a power distribution network as described in any of the foregoing embodiments.

[0203] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0204] The computer instructions stored in the storage medium of the above embodiment are used to cause the computer to execute a method for drone inspection of a power distribution network as described in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be elaborated herein.

[0205] Based on the same concept, corresponding to the method of any of the above embodiments, the present application further provides a computer program product including computer program instructions, which when running on a computer, cause the computer to execute a method for drone inspection of a power distribution network as described in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be elaborated herein.

[0206] It can be understood that before using the technical solutions of the various embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner and the user's authorization will be obtained.

[0207] For example, in response to a user's active request, a prompt message is sent to the user to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the electronic device, application, server, storage medium, or other software or hardware that performs the operation of the disclosed technical solution based on the prompt message.

[0208] As an optional but non-limiting implementation, in response to a user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. Furthermore, the pop-up window may also contain a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0209] It is understandable that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0210] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application is limited to these examples. In line with the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0211] In addition, for simplicity of description and discussion, and in order not to make the embodiment of the application difficult to understand, the known power supply / ground connection with integrated circuit (IC) chip and other components may or may not be shown in the accompanying drawings provided. In addition, the device can be shown in the form of a block diagram to avoid making the embodiment of the application difficult to understand, and this also takes into account the following fact, that is, the details of the embodiment of these block diagram devices are highly dependent on the platform to be implemented in the embodiment of the application (that is, these details should be fully within the scope of understanding of those skilled in the art). When specific details (for example, circuit) are set forth to describe exemplary embodiments of the application, it will be apparent to those skilled in the art that the embodiment of the application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0212] Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may utilize the embodiments discussed.

[0213] The embodiments of the present application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of the present application.

Claims

1. An unmanned aerial vehicle inspection method for a distribution network, characterized in that, Including: Obtaining a three-dimensional model of a distribution network and an inspection task; wherein, the three-dimensional model is pre-constructed according to satellite remote sensing data and unmanned aerial vehicle (UAV) remote sensing data of the distribution network; Determining an inspection path of an inspection UAV based on the inspection task by using the three-dimensional model; Performing an inspection by using the inspection UAV according to the inspection path to obtain an inspection result.

2. The drone inspection method for a distribution network according to claim 1, wherein, The determining the inspection path of the inspection UAV based on the inspection task by using the three-dimensional model includes: Determining an inspection target and the inspection UAV based on the inspection task, and obtaining path planning constraint conditions corresponding to the inspection UAV; Obtaining position information of each inspection target and position information of obstacles by using the three-dimensional model; Determining the inspection path based on the position information of each inspection target, the position information of the obstacles, and the path planning constraint conditions.

3. The method for inspecting a power distribution network by an unmanned aerial vehicle according to claim 2, wherein The path planning constraint conditions include a maximum flight range constraint condition, a minimum step size constraint condition, and a maximum path deflection angle constraint condition; The determining the inspection path based on the position information of each inspection target, the position information of the obstacles, and the path planning constraint conditions includes: Obtaining a preset shooting distance of the inspection target; Determining at least one shooting position information corresponding to each inspection target based on the position information of each inspection target, the preset shooting distance, and the position information of the obstacles; Under the constraints of the maximum flight range constraint condition, the minimum step size constraint condition, and the maximum path deflection angle constraint condition, generating the inspection path by using a shortest path planning optimization algorithm integrated with an obstacle avoidance technology based on each shooting position information and the position information of the obstacles.

4. The method for inspecting a power distribution network by using a drone according to claim 2, wherein, The performing an inspection by using the inspection UAV according to the inspection path to obtain an inspection result includes: Collecting real-time images of the inspection target by using the inspection UAV according to the inspection path to obtain inspection images; Identifying the defect types of the inspection target in the inspection images by using a target detection algorithm and an image classification algorithm to obtain an inspection result.

5. The drone inspection method for a power distribution network according to claim 4, wherein, The identifying the defect types of the inspection target in the inspection images by using a target detection algorithm and an image classification algorithm to obtain an inspection result includes: Determining a target boundary of the inspection target in the inspection image by using the target detection algorithm, and cropping the inspection image based on the target boundary to obtain a target image; Classifying the defect types of the target image by using the image classification algorithm to obtain an inspection result.

6. The method for inspecting a power distribution network by using an unmanned aerial vehicle according to claim 5, characterized in that, The determining the target boundary of the inspection target in the inspection image by using the target detection algorithm includes: Extracting multi-level features of the inspection image by using a multi-scale feature extraction network; Enhancing the multi-level features of the inspection image by using a self-supervised attention mechanism to obtain enhanced features; Fusing the enhanced features with the multi-level features of the inspection image to obtain fused features; Generating candidate bounding boxes based on the fused features by using a region generation network; Outputting the target boundary after performing regression correction and non-maximum suppression operations on the candidate bounding boxes.

7. A method for inspecting a power distribution network by an unmanned aerial vehicle according to claim 5, characterized in that, Using the image classification algorithm to classify the defect types of the target image to obtain the inspection result, including: Extracting the depth features of the target image using a feature extraction network; Inputting the extracted depth features of the target image into a fully connected layer for defect type classification, and outputting the defect type corresponding to the target image.

8. A method for inspecting a power distribution network by an unmanned aerial vehicle according to claim 1, characterized in that, The three-dimensional model is constructed through the following steps: Obtaining the satellite remote sensing data of the power distribution network taken by a low-earth orbit satellite; Using a drone to perform an oblique shot of the power distribution network to obtain the drone remote sensing data; Using image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data to obtain fused image data; Generating the three-dimensional model of the power distribution network based on the fused image data using three-dimensional reconstruction technology.

9. A method for inspecting a power distribution network by an unmanned aerial vehicle according to claim 8, characterized in that, Before using the image fusion technology to fuse the satellite remote sensing data and the drone remote sensing data, it further includes: Performing preprocessing on the satellite remote sensing data, and the preprocessing includes one or several of radiometric calibration, geometric correction, and atmospheric correction.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the program, it implements a drone inspection method for a power distribution network according to any one of claims 1 to 9.

Images