Power transmission line tree obstacle hidden danger automatic identification method based on unmanned aerial vehicle multi-source remote sensing

Abstract

The invention provides a power transmission line tree obstacle hidden danger automatic identification method based on unmanned aerial vehicle multi-source remote sensing. The method comprises the steps of obtaining multi-source remote sensing data; image feature extraction and feature fusion processing are carried out on the high-resolution remote sensing image, and all suspected tree obstacle positions are obtained through detection; classifying the airborne point cloud data, and extracting to obtain a power line result; voxelization classification is carried out on the airborne point cloud data in the coverage area where the target suspected tree obstacle position is located, and a tree classification result is obtained; according to the tree classification result and the power line result, calculating the distance between the tree and the power line in the coverage range to carry out gradient division to obtain a division result; and under the condition that the division result is the tree obstacle area, determining the tree obstacle hidden danger level of the tree obstacle area where the target suspected tree obstacle position is located, and judging all the suspected tree obstacle positions to obtain all the tree obstacle areas and the corresponding tree obstacle hidden danger levels. According to the method, the problem of low inspection efficiency of the tree obstacles of the power transmission line in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line tree barrier detection, and in particular to a method for automatically identifying power transmission line tree barrier hazards based on unmanned aerial vehicle multi-source remote sensing, a device for automatically identifying power transmission line tree barrier hazards based on unmanned aerial vehicle multi-source remote sensing, a computer-readable storage medium, and a system for automatically identifying power transmission line tree barrier hazards. Background Art

[0002] Transmission lines are key infrastructure of modern power systems. They bear the heavy responsibility of transmitting electricity over long distances, delivering electricity generated by power plants to end users in cities, industrial areas, and rural areas. With the growth of global electricity demand and the continuous expansion of the scale of power networks, the coverage of transmission lines has gradually extended to more complex geographical areas. These lines often cross different terrains such as cities, villages, forests, mountains, and lakes. The complexity of the external natural environment poses severe challenges to the daily operation and maintenance and safety of transmission lines, especially in areas with many trees such as forests and rural areas. The vegetation around the transmission lines poses a potential threat to the safety of the lines. The growth of trees is natural and uncontrollable. Over time, trees or branches may come into contact with power lines, especially in extreme weather conditions such as strong winds, heavy rains, or ice and snow. This is more likely to happen. Once branches or trees come into contact with wires, accidents such as short circuits, fires, or power outages may occur, resulting in power outages, and in severe cases, even large-scale power grid failures. This will not only affect residential electricity consumption, but also cause huge losses to industrial production and social operations. Therefore, ensuring the safe and stable operation of transmission lines is a vital task for power grid operators.

[0003] In order to prevent potential safety hazards to the power grid caused by excessive tree growth, it is usually necessary to regularly inspect and evaluate tree barriers around transmission lines to ensure that the distance between the lines and surrounding trees remains within a safe range. Once it is found that the growth of trees threatens the safety of the lines, timely measures need to be taken, such as pruning or removing trees, to prevent accidents caused by tree barriers.

[0004] At present, in actual operation, the detection of tree obstacle hazards usually relies on manual inspections. Manual inspections require inspectors to walk along the transmission lines or use transportation to conduct on-site inspections, which is time-consuming, labor-intensive, and inefficient. Since transmission lines often pass through areas with complex terrain, manual inspections also pose great safety risks. For example, inspectors need to work in steep mountains, dense forests, or high-risk areas, which not only consumes a lot of time and manpower costs, but also may face safety threats in the working environment. In addition, manual inspections rely on the experience and subjective judgment of inspectors, and the accuracy and consistency of the inspection results are difficult to guarantee, and it is easy to miss inspections or misjudge.

[0005] In order to solve these problems, drone technology has been gradually introduced into the inspection of power transmission lines in recent years. Drones can be equipped with high-resolution cameras and sensing equipment such as LiDAR, flying over power transmission lines to conduct comprehensive monitoring of the lines and their surroundings. Through the data collected by drones, high-definition images and three-dimensional point cloud data around the lines can be quickly obtained, thereby improving inspection efficiency and coverage. However, at this stage, the images and point cloud data of drone inspections usually still require post-manual analysis, relying on professionals to identify and analyze tree obstacles, and cannot achieve fully automated detection. There are still problems such as low efficiency and subjective results. Summary of the invention

[0006] The main purpose of the present application is to provide a method for automatically identifying tree barrier hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles, an automatic identification device for tree barrier hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles, a computer-readable storage medium and an automatic identification system for tree barrier hazards in power transmission lines, so as to at least solve the problem of low inspection efficiency of tree barriers in power transmission lines in the prior art.

[0007] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles is provided, comprising: obtaining multi-source remote sensing data, wherein the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, wherein the high-resolution remote sensing images are remote sensing images of power lines of the power transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers; using the high-resolution remote sensing images to perform image feature extraction and feature fusion processing based on an improved target detection model to detect all suspected tree obstacle locations, wherein the improved target detection model is based on YOLOv8 An improved model is obtained by introducing a channel attention mechanism on the basis of the model; the airborne point cloud data is classified based on the improved point cloud deep learning model to extract the power line result, and the improved point cloud deep learning model is an improved model obtained by adding a special classifier on the output layer of the PointNet model, and the special classifier is a classifier dedicated to separating linear point cloud data; a retrieval and classification step is performed to retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize and classify the target point cloud data to obtain The tree classification result, the target suspected tree barrier position is any one of the suspected tree barrier positions, and the point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as an activation function based on the VoxelNet model; a calculation step, according to the tree classification result and the power line result, calculating the distance between the trees and the power lines within the coverage area, and performing gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance. an area at a safe distance; a determination step, when the division result is the tree barrier area, analyzing according to the corresponding distance in combination with the power of the transmission line, to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, the tree barrier hazard level being one of the following: a high hazard level, a medium hazard level and a low hazard level; repeating the retrieval classification step, the calculation step and the determination step at least once in sequence until all the suspected tree barrier positions are judged, to obtain all the tree barrier areas and the corresponding tree barrier hazard levels, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels.

[0008] Optionally, multi-source remote sensing data is obtained, including: planning the UAV flight trajectory according to the direction and location distribution of the power lines to obtain the target UAV flight trajectory; controlling the medium and high resolution remote sensing equipment carried by the UAV platform to obtain fine image data of the power lines according to the target UAV flight trajectory, and at the same time, using the laser radar carried by the UAV platform to obtain the airborne point cloud data.

[0009] Optionally, the improved target detection model includes a Backbone module, a Neck module and a Head module, the Backbone module is used to extract image features and generate a multi-scale feature map, the Neck module is used to perform feature fusion processing on the multi-scale feature map to generate a fused feature map, the Head module is used to detect and output the fused feature map, and the high-resolution remote sensing image is used to extract image features and perform feature fusion processing based on the improved target detection model to detect all suspected tree barrier positions, including: sequentially performing image resizing, normalization processing and image enhancement processing on the high-resolution remote sensing image to obtain a target resolution remote sensing image; inputting the target resolution remote sensing image into the improved target detection model, using the Backbone module in the improved target detection model to extract image features of the target resolution remote sensing image through multi-layer convolution operations to generate the multi-scale feature map, using the Neck module to fuse the multi-scale feature map through a feature pyramid network and a path aggregation network to obtain a fused feature map, and using the Head module to detect the fused feature map and output all the suspected tree barrier positions in sequence to obtain all the suspected tree barrier positions.

[0010] Optionally, the network structure of the improved point cloud deep learning model includes a transformation network, a feature extraction network, and a global feature aggregation and classification output module, the global feature aggregation and classification output module includes at least a maximum pooling layer and the output layer, and the airborne point cloud data is classified based on the improved point cloud deep learning model to extract the power line result, including: performing removal operations and normalization processing on the airborne point cloud data in sequence to obtain valid airborne point cloud data, the removal operation is an operation to remove point cloud data that is not related to the power line; the valid airborne point cloud data is input into the improved point cloud deep learning model, so that the transformation network in the improved point cloud deep learning model aligns the valid airborne point cloud data to a standard transformation matrix through multiple convolutional layers and fully connected layers, the feature extraction network uses a multilayer perceptron to extract local point features from the transformation matrix, the maximum pooling layer in the global feature aggregation and classification module aggregates the local point features into a global feature vector, and remaps the global feature vector through the special classifier in the output layer to generate linear point cloud features for classification output to obtain the power line result.

[0011] Optionally, the network structure of the improved point cloud object detection model includes a voxelization module, a feature extraction module and a classification and regression module, the voxelization module is used to perform voxelization operations, the feature extraction module includes a voxel feature encoding module and a convolutional intermediate layer, the classification and regression module is used to classify and locate the target object through a region proposal network, retrieve the airborne point cloud data within the coverage range of the target suspected tree barrier position, obtain the target point cloud data corresponding to the target suspected tree barrier position, and use the improved point cloud object detection model to voxelize and classify the target point cloud data to obtain a tree classification result, including: using the improved target detection model to retrieve a suspected tree barrier detection box corresponding to the target suspected tree barrier position; positioning the suspected tree barrier detection box in the image coordinate system; The target position is mapped into a three-dimensional space coordinate system, and a corresponding three-dimensional search area is set; the airborne point cloud data falling within the three-dimensional search area is screened out to obtain the target point cloud data; the target point cloud data is input into the improved point cloud object detection model, so that the voxelization module in the improved point cloud object detection model divides the target point cloud data into a plurality of voxel grids of set sizes, the voxel feature encoding module performs local feature extraction on the target point cloud data in the non-empty voxel grids to obtain voxel features, and the convolutional intermediate layer further extracts the spatial features between voxels for all the voxel features, and the region proposal network classifies and locates the trees in the three-dimensional search area according to the spatial features to obtain the tree classification result.

[0012] Optionally, according to the tree classification result and the power line result, the distance between the trees and the power line within the coverage area is calculated, and gradient division is performed according to the distance to obtain the division result, including: calculating the distance according to a first formula, wherein the first formula is d represents the distance between the tree and the power line, v represents the direction vector of the power line, p0 represents the coordinates of any coordinate point on the power line, and q represents the coordinates of any tree in the voxelized area; the distance is compared with the first tree barrier safety distance; when the distance is less than or equal to the first tree barrier safety distance, the division result is determined to be the tree barrier area; when the distance is greater than the first tree barrier safety distance, the division result is determined to be the non-tree barrier area.

[0013] Optionally, when the division result is the tree barrier area, an analysis is performed according to the corresponding distance in combination with the power of the transmission line to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, including: determining the minimum tree barrier safety distance between the tree and the transmission line according to the power of the transmission line; when the distance is less than or equal to the first tree barrier safety distance and greater than the second tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the low hazard level, and the second tree barrier safety distance is less than the first tree barrier safety distance and greater than the minimum tree barrier safety distance; when the distance is less than or equal to the second tree barrier safety distance and greater than the minimum tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the medium hazard level; when the distance is less than or equal to the minimum tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the high hazard level.

[0014] According to another aspect of the present application, there is provided an automatic identification device for tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles, the device comprising: an acquisition unit for acquiring multi-source remote sensing data, the multi-source remote sensing data comprising at least high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images being remote sensing images of power lines of the power transmission lines, the airborne point cloud data comprising at least spatial positioning data of the power lines, trees and transmission towers; a feature processing unit for performing image feature extraction and feature fusion processing based on an improved target detection model using the high-resolution remote sensing images to detect all suspected tree obstacle locations, the improved target detection model being a YOLOv8 model An improved model is obtained by introducing a channel attention mechanism on the basis of ; an extraction unit is used to classify the airborne point cloud data based on the improved point cloud deep learning model, and extract the power line result. The improved point cloud deep learning model is an improved model obtained by adding a special classifier to the output layer of the PointNet model, and the special classifier is a classifier dedicated to separating linear point cloud data; a retrieval classification unit is used to perform a retrieval classification step, retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize and classify the target point cloud data. A tree classification result is obtained, the target suspected tree barrier position is any one of the suspected tree barrier positions, and the point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as an activation function based on the VoxelNet model; a computing unit is used to execute a computing step, calculate the distance between the trees and the power lines within the coverage area according to the tree classification result and the power line result, and perform gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance a determining unit, configured to execute a determining step, and when the division result is the tree barrier area, analyzing according to the corresponding distance in combination with the power of the transmission line, to determine a tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, wherein the tree barrier hazard level is one of the following: a high hazard level, a medium hazard level, and a low hazard level; and a repeating unit, configured to sequentially repeat the retrieval and classification step, the calculation step, and the determination step at least once until all the suspected tree barrier positions are judged, so as to obtain all the tree barrier areas and the corresponding tree barrier hazard levels, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels.

[0015] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0016] According to another aspect of the present application, there is provided a system for automatically identifying tree obstacles in power transmission lines, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the methods described.

[0017] By applying the technical solution of the present application, in the automatic identification method of tree obstacle hidden dangers in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles, first, multi-source remote sensing data is obtained, and the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data. The high-resolution remote sensing image is a remote sensing image of the power line of the transmission line, and the airborne point cloud data at least includes spatial positioning data of power lines, trees and transmission towers; then, the high-resolution remote sensing image is used to extract image features and perform feature fusion processing based on the improved target detection model to detect all suspected tree obstacle locations. The improved target detection model is based on the YOLOv8 model. An improved model is obtained by introducing the channel attention mechanism on the basis of the proposed method; then, the airborne point cloud data is classified based on the improved point cloud deep learning model to extract the power line results. The improved point cloud deep learning model is an improved model obtained by adding a dedicated classifier to the output layer of the PointNet model. The dedicated classifier is a classifier dedicated to separating linear point cloud data; then, the retrieval and classification steps are performed to retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use the improved point cloud object detection model to perform object detection on the target point cloud data. voxel classification is performed to obtain a tree classification result. The target suspected tree barrier position is any suspected tree barrier position. The point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model. Then, a calculation step is performed to calculate the distance between the trees and the power lines within the coverage area according to the tree classification result and the power line result, and gradient division is performed according to the distance to obtain a division result. The division result is a tree barrier area or a non-tree barrier area. The tree barrier area is an area with a distance less than the first tree barrier safety distance, and the non-tree barrier area is an area with a distance greater than the first tree barrier safety distance. Then, a determination step is performed. When the division result is a tree barrier area, the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located is determined according to the corresponding distance combined with the power of the transmission line. The tree barrier hazard level is one of the following: high hazard level, medium hazard level and low hazard level. Finally, the retrieval classification step, the calculation step and the determination step are repeated at least once in sequence until all suspected tree barrier positions are judged, and all tree barrier areas and the corresponding tree barrier hazard levels are obtained, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels. This application fully utilizes the advantages of image information and point cloud data collected by drones and processes them directly. Based on the improved YOLO v8 model, image features are extracted and fused from the acquired high-resolution remote sensing images to detect and obtain the locations of suspected tree barriers in order to further analyze potential risk areas; then, the improved PointNet model is used to classify the point cloud data, extract the power lines, and retrieve the point cloud data in the area based on the detected suspected tree barrier locations.Subsequently, the improved VoxelNet model is used to classify these point cloud data, thereby identifying and voxelizing the tree areas to provide support for subsequent precise analysis; finally, the distance between the trees and the power lines is calculated based on the classification results, and the distance calculation results are divided to exclude non-tree barrier areas, thereby achieving accurate interpretation of the area. Combined with the power of the transmission line, the tree barrier hazard level is determined, and the final detection results are output and displayed. Compared with the existing methods, the method of the present application can obtain tree barrier extraction results more quickly and accurately. The present application solves the problem of low efficiency in inspection of tree barriers on transmission lines in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A hardware structure block diagram of a mobile terminal for executing a method for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles provided in an embodiment of the present application is shown;

[0019] Figure 2 A schematic flow chart of a method for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing using drones according to an embodiment of the present application is shown;

[0020] Figure 3 A schematic diagram of a flow chart of a method for automatically identifying tree obstacle hazards in power transmission lines according to an embodiment of the present application is shown;

[0021] Figure 4 A framework diagram of an improved point cloud deep learning model provided according to an embodiment of the present application is shown;

[0022] Figure 5 A structural block diagram of a device for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing by unmanned aerial vehicles is shown according to an embodiment of the present application.

[0023] The above drawings include the following reference numerals:

[0024] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0025] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0029] Tree barriers on power transmission lines: As trees grow, the distance between them and power lines decreases, causing electric field distortion on the surface of the power lines. The intensity of the electric field between the power lines and the trees increases, eventually leading to flashover of the transmission lines and causing power outages.

[0030] Backbone, also known as the backbone network, is the most basic part of the deep learning model. Its main task is to extract the features of the input data. In image processing tasks, Backbone is usually a convolutional neural network (CNN) that is used to extract useful information from the original image, such as edges, textures, shapes, etc. The performance of the Backbone directly affects the performance of the entire model, so it is very important to choose a suitable Backbone. Common Backbone networks include ResNet, VGG, MobileNet, etc. These network structures have been trained and optimized a lot and have been proven to perform well in a variety of tasks.

[0031] Neck module: Located between Backbone and Head, its main task is to further integrate and process the features extracted by Backbone. In tasks such as object detection and face recognition, the Neck module usually includes some specific layers, such as ROI Pooling, FPN (Feature Pyramid Network), etc. These layers can effectively integrate features at different levels and improve the performance of the model. In addition, the Neck module can also introduce an attention mechanism to make the model pay more attention to important features.

[0032] Head module: The last layer of the deep learning model, responsible for mapping the features processed by Backbone and Neck to the final output space. Depending on the task, the structure of the Head module will also be different. For example, in the image classification task, the Head module is usually a fully connected layer or a softmax classifier; in the target detection task, the Head module may include a bounding box regressor and a classifier, etc. As introduced in the background technology, the images and point cloud data of drone inspections in the prior art at this stage usually still require later manual analysis, relying on professionals to identify and analyze tree barrier hazards, and cannot achieve fully automated detection. There are still problems such as low efficiency and subjective results. In order to solve the problem of low efficiency in the inspection of transmission line tree barriers, the embodiments of the present application provide a method for automatically identifying transmission line tree barrier hazards based on drone multi-source remote sensing, a device for automatically identifying transmission line tree barrier hazards based on drone multi-source remote sensing, a computer-readable storage medium, and a system for automatically identifying transmission line tree barrier hazards.

[0033] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0034] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 1 is a hardware structure block diagram of a mobile terminal of an automatic identification method of tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0035] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the automatic identification method of tree obstacles in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific examples of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0036] In this embodiment, a method for automatically identifying tree obstacles in power transmission lines based on multi-source remote sensing by unmanned aerial vehicles, which runs on a mobile terminal, a computer terminal or a similar computing device, is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0037] Figure 2 1 is a flow chart of a method for automatically identifying tree obstacles in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0038] Step S201, obtaining multi-source remote sensing data, the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images are remote sensing images of power lines of transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers.

[0039] Specifically, when planning the route of a drone, it is necessary to comprehensively consider the direction and location of power lines to ensure the safety and efficiency of the route. Multi-source remote sensing refers to a technology that uses remote sensing data from different sensors, different time periods or different spatial resolutions for comprehensive analysis and processing. In the drone platform, it includes high-resolution images and LiDAR point cloud data. The drone route is planned according to the direction and location of the power lines, and the drone multi-sensor platform is used to obtain multi-source remote sensing data, including high-resolution remote sensing images and airborne LiDAR point cloud data.

[0040] Step S202, using the above-mentioned high-resolution remote sensing image to perform image feature extraction and feature fusion processing based on the improved target detection model, to detect all suspected tree obstacle locations, wherein the above-mentioned improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of the YOLOv8 model.

[0041] Specifically, the high-resolution remote sensing images are used to extract and fuse image features based on the improved YOLO v8 model to detect and obtain the suspected tree obstacle locations. In the feature fusion process of the YOLO v8 model, the SENet channel attention mechanism is introduced to improve the expression ability of tree obstacle features by adaptively adjusting the weight of each channel, reduce the impact of low-level vegetation on detection, and thus improve the model detection performance.

[0042] Step S203, classify the above-mentioned airborne point cloud data based on the improved point cloud deep learning model to extract the power line results. The above-mentioned improved point cloud deep learning model is an improved model obtained by adding a special classifier to the output layer of the PointNet model. The above-mentioned special classifier is a classifier dedicated to separating linear point cloud data.

[0043] Specifically, on the basis of the original PointNet model, a special classifier is added to the output layer, LineSegNet (Line Segment Network), which is a classifier dedicated to separating linear point cloud structures. It is called the PointNet-PL model (i.e., the above-mentioned improved point cloud deep learning model). This model can further optimize the network's ability to separate linear point clouds, and can simultaneously combine geometric features and contextual information to finely classify power lines in point clouds; in addition, a multi-task learning framework is adopted to enhance the classifier's attention to power lines through a weighted loss function to ensure the accuracy of the network in identifying power lines. Through this improvement, the above-mentioned airborne point cloud data is input into the above-mentioned improved point cloud deep learning model for classification, and the power line results are extracted, which can more accurately extract power line features and effectively improve the high-precision classification of power line point clouds.

[0044] Step S204, a retrieval and classification step, retrieves the above-mentioned airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtains the target point cloud data corresponding to the above-mentioned target suspected tree obstacle position, and uses the improved point cloud object detection model to voxelize and classify the above-mentioned target point cloud data to obtain a tree classification result. The above-mentioned target suspected tree obstacle position is any one of the above-mentioned suspected tree obstacle positions, and the above-mentioned point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model.

[0045] Specifically, the suspected tree obstacle detection frame obtained by the YOLO v8 network is used to map the position of the detection frame in the image coordinate system to the three-dimensional space coordinate system, and a corresponding three-dimensional retrieval area is defined to filter out the point cloud falling in the area from the original point cloud data, that is, the above-mentioned target point cloud data. In the VoxelNet model, the GELU activation function is used to replace the traditional ReLU function to increase the feature expression ability of the network structure, so that it can obtain accurate tree voxels based on the tree structure characteristics more delicately, and it is called the VoxelNet-TR model (that is, the above-mentioned improved point cloud object detection model). The above-mentioned target point cloud data is input into the VoxelNet-TR model for voxel classification to obtain the tree classification result.

[0046] It should be noted that the present invention does not calculate all point cloud data, which reduces the amount of calculation and improves model efficiency. The PointNet-PL model (i.e., the above-mentioned improved point cloud deep learning model) is used for power line extraction, and the VoxelNet-TR model (i.e., the above-mentioned improved point cloud object detection model) performs voxel classification on the tree barrier point cloud, achieving accurate separation and identification of tree barriers and power lines, avoiding the calculation of all point cloud data, reducing the model's dependence on computer performance, and improving calculation efficiency.

[0047] Step S205, a calculation step, calculates the distance between the trees and the power lines within the coverage area according to the tree classification results and the power line results, and performs gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance.

[0048] Specifically, the distance between the power lines and the tree voxel grids is calculated using the classified power lines and voxelized trees, and the calculation results are divided to obtain the division results so as to exclude the non-tree barrier areas.

[0049] Step S206, a determination step, in the case where the above-mentioned division result is the above-mentioned tree barrier area, an analysis is performed based on the corresponding above-mentioned distance combined with the power of the above-mentioned transmission line to determine the tree barrier hazard level of the above-mentioned tree barrier area where the above-mentioned target suspected tree barrier position is located, and the above-mentioned tree barrier hazard level is one of the following: a high hazard level, a medium hazard level and a low hazard level.

[0050] Specifically, based on the determination of the tree barrier area, combined with the power level of the transmission line, the tree barrier hazards and the potential impact of tree barriers on the operation of the transmission line are further evaluated. The power of the transmission line determines its ability to transmit electrical energy and the magnitude of the current. Lines with higher power usually have stricter safety distance requirements for tree barriers. For tree barriers within the line protection area, the tree barrier hazard level is determined based on the distance outside the edge line determined in the "Overhead Transmission Line Operation Regulations".

[0051] Step S207, repeatedly executing the above-mentioned retrieval classification step, the above-mentioned calculation step and the above-mentioned determination step at least once in sequence until the judgment of all the above-mentioned suspected tree barrier positions is completed, and all the above-mentioned tree barrier areas and the corresponding above-mentioned tree barrier hazard levels are obtained, so as to inspect the corresponding above-mentioned tree barrier areas according to the above-mentioned tree barrier hazard levels.

[0052] Specifically, the above-mentioned retrieval and classification step, the above-mentioned calculation step and the above-mentioned determination step are iteratively executed until the interpretation of all suspected tree barrier locations is completed, all non-tree barrier areas are gradually excluded, and the real tree barrier areas and the corresponding hidden danger levels are graded and displayed, so that the corresponding tree barrier areas can be inspected according to the above-mentioned tree barrier hidden danger levels to ensure the accuracy and comprehensiveness of the inspection results.

[0053] In this embodiment, first, multi-source remote sensing data is acquired, and the multi-source remote sensing data includes at least high-resolution remote sensing images and airborne point cloud data. The high-resolution remote sensing image is a remote sensing image of the power line of the transmission line, and the airborne point cloud data includes at least spatial positioning data of the power line, trees and transmission towers; then, the high-resolution remote sensing image is used to perform image feature extraction and feature fusion processing based on the improved target detection model to detect all suspected tree obstacle positions, and the improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of the YOLOv8 model; then, the airborne point cloud data is classified based on the improved point cloud deep learning model to extract the power line results, and the improved point cloud deep learning model is an improved model obtained by adding a dedicated classifier to the output layer of the PointNet model, and the dedicated classifier is a classifier dedicated to separating linear point cloud data; then, the retrieval and classification steps are performed to retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize the target point cloud data to obtain the tree classification result. The target suspected tree barrier position is any suspected tree barrier position, and the point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model; then, the calculation step is executed, and the distance between the trees and the power lines within the coverage range is calculated according to the tree classification results and the power line results, and gradient division is performed according to the distance to obtain the division result, and the division result is a tree barrier area or a non-tree barrier area. The tree barrier area is an area with a distance less than the first tree barrier safety distance, and the non-tree barrier area is an area with a distance greater than the first tree barrier safety distance; then, the determination step is executed. In the case where the division result is a tree barrier area, the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located is determined according to the corresponding distance combined with the power of the transmission line. The tree barrier hazard level is one of the following: high hazard level, medium hazard level and low hazard level; finally, the retrieval classification step, the calculation step and the determination step are repeated at least once in sequence until all suspected tree barrier positions are judged, and all tree barrier areas and the corresponding tree barrier hazard levels are obtained, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels. This application fully utilizes the advantages of image information and point cloud data collected by drones and processes them directly. Based on the improved YOLO v8 model, image features are extracted and fused from the acquired high-resolution remote sensing images to detect and obtain the locations of suspected tree barriers in order to further analyze potential risk areas; then, the improved PointNet model is used to classify the point cloud data, extract the power lines, and retrieve the point cloud data in the area based on the detected suspected tree barrier locations.Subsequently, the improved VoxelNet model is used to classify these point cloud data, thereby identifying and voxelizing the tree areas to provide support for subsequent precise analysis; finally, the distance between the trees and the power lines is calculated based on the classification results, and the distance calculation results are divided to exclude non-tree barrier areas, thereby achieving accurate interpretation of the area. Combined with the power of the transmission line, the tree barrier hazard level is determined, and the final detection results are output and displayed. Compared with the existing methods, the method of the present application can obtain tree barrier extraction results more quickly and accurately. The present application solves the problem of low efficiency in inspection of tree barriers on transmission lines in the prior art.

[0054] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for automatic identification of tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles of the present application will be described in detail below in combination with specific embodiments.

[0055] In order to improve the accuracy and comprehensiveness of tree obstacle detection, in an optional implementation manner, the above step S201 includes:

[0056] Step S2011, planning the flight trajectory of the UAV according to the direction and position distribution of the power lines to obtain the flight trajectory of the target UAV;

[0057] Step S2012, according to the flight trajectory of the target UAV, control the medium and high resolution remote sensing equipment carried by the UAV platform to obtain fine image data of the power lines, and at the same time, use the laser radar carried by the above UAV platform to obtain the above airborne point cloud data.

[0058] In the above embodiments, the flight trajectory of the UAV is manually planned according to the direction and location distribution of the transmission line to ensure that the flight path can fully cover the key areas of the transmission line and its surrounding environment; by setting the appropriate flight altitude, speed and route of the UAV, the high-resolution remote sensing equipment carried by the UAV platform is used to obtain fine image data of the power line, accurately capture the details of the line and the changes in the surrounding environment. The multispectral sensor equipment obtains high-resolution image data to capture the details of the transmission line and the changes in the surrounding environment; at the same time, the laser radar system carried by the UAV platform is used to obtain three-dimensional LiDAR point cloud data to achieve spatial positioning of different targets in the transmission line (such as power lines, trees, transmission towers, etc.), which can effectively cover the transmission line in different environments, and improve the accuracy and comprehensiveness of the detection. In this example, high-resolution remote sensing images are obtained by using optical sensors for ground observation and collection; airborne laser radar point cloud data are obtained by using airborne laser radar sensors for ground observation and collection.

[0059] In order to improve the expression capability of tree obstacle features to accurately obtain the suspected tree obstacle location, in an optional implementation manner, the above step S202 includes:

[0060] Step S2021, sequentially performing image resizing, normalization processing, and image enhancement processing on the high-resolution remote sensing image to obtain a remote sensing image of target resolution;

[0061] Step S2022, input the above-mentioned target resolution remote sensing image into the above-mentioned improved target detection model, use the above-mentioned Backbone module in the above-mentioned improved target detection model to extract the image features of the above-mentioned target resolution remote sensing image through multi-layer convolution operations to generate the above-mentioned multi-scale feature map, use the above-mentioned Neck module to fuse the above-mentioned multi-scale feature map through a feature pyramid network and a path aggregation network to obtain a fused feature map, use the above-mentioned Head module to detect the above-mentioned fused feature map and output all the above-mentioned suspected tree obstacle positions in sequence to obtain all the suspected tree obstacle positions.

[0062] In the above embodiment, data preprocessing is first performed to normalize and enhance the high-resolution remote sensing images to improve the robustness and generalization ability of the model. Figure 3As shown in the upper part, the target high-resolution remote sensing image is input into the improved YOLO v8 model. The model structure of YOLO v8 mainly includes three parts: Backbone, Neck and Head. Each part plays a key role in feature extraction, fusion and target detection. The Backbone module is used to extract image features and generate multi-scale feature maps. The Neck module is used to perform feature fusion processing on the multi-scale feature maps to generate fused feature maps. The Head module is used to detect the fused feature maps and output them. The preprocessed image (the target resolution remote sensing image) is input into the Backbone module for feature extraction. The core of this stage is to extract the low-level to high-level feature maps of the image through multi-layer convolution operations. Backbone uses a deep neural network to ensure that key visual information can be fully expressed. The Backbone module mainly extracts individual scale features from the images generated by the C2f module and the SPPF module for feature extraction. The C2f module replaces the original C3 module. After reducing one convolution layer, the model is more lightweight. SPPF reduces the number of network layers on the basis of the original SPP, which can perform feature fusion faster and eliminate redundant operations. Next, the extracted multi-scale feature maps are passed to the Neck module. The Neck module can effectively fuse feature maps of different resolutions and semantic levels through the combination of feature pyramid network (FPN) and path aggregation network (PAN). FPN enhances the semantic expression ability of higher-level feature maps through top-down feature fusion, ensuring that the model can understand complex objects in the image. At the same time, PAN retains the detailed information of lower-level feature maps through bottom-up path enhancement, making the model more accurate when detecting targets of different sizes. Through the introduction of SENet channel attention mechanism, the weight of each channel can be adaptively adjusted, and the global information of each channel is extracted by global average pooling. Then, weights are generated through two layers of fully connected networks. After the weight distribution between channels, the network can focus on more meaningful channel features and suppress irrelevant or redundant information. This mechanism enhances the model's ability to capture important features and improves the accuracy of target detection, especially when dealing with complex scenes and multi-scale targets. After feature fusion is completed, the output of the Neck module is passed to the Head module, which is responsible for converting the fused feature map into a specific detection result. Through a series of convolution operations and classifiers, the model can output the specific location and category information of the tree barrier. In the feature fusion process of the YOLO v8 model, the SENet channel attention mechanism is introduced to improve the expression ability of tree barrier features by adaptively adjusting the weight of each channel, reduce the impact of low-level vegetation on detection, and thus improve the model detection performance.

[0063] In order to more accurately extract the power line features and effectively improve the high-precision classification of the power line point cloud, in an optional implementation, the above step S203 includes:

[0064] Step S2031, performing removal operation and normalization processing on the above-mentioned airborne point cloud data in sequence to obtain valid airborne point cloud data, wherein the above-mentioned removal operation is an operation of removing point cloud data irrelevant to the power line;

[0065] Step S2032, input the above-mentioned valid airborne point cloud data into the above-mentioned improved point cloud deep learning model, so that the above-mentioned transformation network in the above-mentioned improved point cloud deep learning model aligns the above-mentioned valid airborne point cloud data to the standard transformation matrix through multiple convolutional layers and fully connected layers, the above-mentioned feature extraction network uses a multi-layer perceptron to extract local point features from the above-mentioned transformation matrix, the maximum pooling layer in the above-mentioned global feature aggregation and classification module aggregates the above-mentioned local point features into a global feature vector, and remaps the above-mentioned global feature vector through the above-mentioned special classifier in the above-mentioned output layer to generate linear point cloud features for classification output, so as to obtain the above-mentioned power line result.

[0066] In the above embodiment, the network structure of the above improved point cloud deep learning model includes a transformation network, a feature extraction network and a global feature aggregation and classification output module, and the above global feature aggregation and classification output module includes at least a maximum pooling layer and the above output layer. Remove irrelevant point cloud data such as ground, building and other noise data from the collected airborne point cloud data to reduce the impact of errors on classification, and normalize the point cloud data so that the spatial positions of all points are standardized to a unified coordinate range to obtain the above valid airborne point cloud data. Figure 4 As shown in the figure, first, the point cloud data (that is, the valid airborne point cloud data) is input into the transformation network using the T-Net structure. Through multiple 3×3 convolutional layers and fully connected layers, the transformation matrix that aligns the input valid airborne point cloud data to the standard posture is learned. The feature extraction network uses point-by-point MLP (multi-layer perceptron) to extract point features, where each layer of MLP is equivalent to a 1×1 convolutional layer to extract the local features of the point, and the activation function is ReLU. In this process, no spatial pooling operation is performed to maintain the spatial information of the point. The feature vector output by the feature extraction network is aggregated into a global feature vector through the maximum pooling operation. This operation integrates the local information of the point cloud into a global descriptor of a fixed length, which is invariant to spatial transformation. Finally, through several fully connected layers, the network maps the global features to the output space of a specific task (such as classification or segmentation), and finally calculates the category probability through the Softmax layer or realizes point cloud segmentation through the point-by-point MLP.

[0067] In this embodiment, if Figure 4As shown in the figure, the PointNet model is improved as follows: Based on the original PointNet model, a special classifier is added to the output layer, LineSegNet (Line Segment Network) is a classifier dedicated to separating linear point cloud structures; in addition, a multi-task learning framework is adopted to enhance the classifier's attention to power lines through a weighted loss function to ensure the accuracy of the network in identifying power lines. The improved model is called the PointNet-PL model (i.e. the above-mentioned improved point cloud deep learning model). Based on the PointNet-PL model, the power lines are extracted by classifying the point cloud data. The LineSegNet classifier is added to the feature aggregation module of the original PointNet. The classifier remaps the aggregated global features through a dedicated MLP layer to generate specific features of the linear point cloud. In this process, the structure of the MLP layer is consistent with the original PointNet, but an additional fully connected layer with specific weight initialization is added before the output layer to enhance the recognition of linear structures. In the output layer of the classifier, a weighted loss function is introduced to give a higher weight specifically to the power line category, thereby enhancing the classifier's attention to the power line. Through this improvement, the PointNet-PL model can extract power line features more accurately and effectively improve the high-precision classification of power line point clouds.

[0068] In order to ensure that tree information can be detected more accurately in complex scenes, in an optional implementation, the above step S204 includes:

[0069] Step S2041, using the improved target detection model to retrieve a suspected tree obstacle detection frame corresponding to the suspected tree obstacle position of the target;

[0070] Step S2042, mapping the coordinate position of the suspected tree obstacle detection frame in the image coordinate system to the three-dimensional space coordinate system, and setting a corresponding three-dimensional search area;

[0071] Step S2043, filtering out the airborne point cloud data falling within the three-dimensional search area to obtain the target point cloud data;

[0072] Step S2044, input the above-mentioned target point cloud data into the above-mentioned improved point cloud object detection model, so that the above-mentioned voxelization module in the above-mentioned improved point cloud object detection model divides the above-mentioned target point cloud data into multiple voxel grids of set sizes, and uses the above-mentioned voxel feature encoding module to perform local feature extraction on the above-mentioned target point cloud data in the above-mentioned non-empty voxel grids to obtain voxel features, and further extracts the spatial features between voxels for all the above-mentioned voxel features through the above-mentioned convolutional intermediate layer, and the above-mentioned region proposal network classifies and locates the trees in the above-mentioned three-dimensional retrieval area according to the above-mentioned spatial features to obtain the above-mentioned tree classification results.

[0073] In the above embodiment, the network structure of the improved point cloud object detection model includes a voxelization module, a feature extraction module and a classification regression module. The voxelization module is used to perform voxelization operations, the feature extraction module includes a voxel feature encoding module and a convolutional intermediate layer, and the classification regression module is used to classify and locate the target object through a region proposal network. First, the suspected tree obstacle detection frame retrieved by the improved YOLO v8 model is mapped to the position of the suspected tree obstacle detection frame in the image coordinate system into the three-dimensional space coordinate system, and a corresponding three-dimensional retrieval area is defined to filter out the point cloud (i.e., the target point cloud data) falling within the area from the original point cloud data. The target point cloud data is divided into a three-dimensional voxel grid of a fixed size by the improved VoxelNet model. The point cloud in each non-empty voxel is processed by the VFE module to extract local features. The VFE module uses a multi-layer perceptron (MLP) with shared parameters to perform feature mapping on the coordinates of each point, cancels the traditional ReLU function, and applies the GELU activation function. This process can be expressed as: ij =GELU(W 1 p ij +b 1 ), where p ij Represents voxel v j The coordinates of the i-th point in W 1 and b 1are the weight matrix and bias term of MLP respectively. The network is downsampled through the maximum pooling layer after the convolution operation, and the spatial size of the feature map is reduced layer by layer. The second and third layers use 128 and 256 convolution kernels for feature extraction, respectively. The spatial size of the final feature map is reduced to 1 / 8 of the original voxel grid, but the number of channels is increased to 256. On the feature map extracted by 3D convolution, the region proposal network (RPN) classifies and locates the target object. RPN contains a 3D convolution layer, which is then divided into a classification branch and a regression branch. The former calculates the category probability through Softmax, and the latter outputs the bounding box regression parameters. The entire network structure uses the GELU activation function to improve the nonlinear expression ability, so that the improved VoxelNet model (also known as the VoxelNet-TR model) can more accurately detect tree information in complex scenes.

[0074] In order to accurately exclude non-tree barrier areas, in an optional implementation manner, the above step S205 includes:

[0075] Step S2051, calculating the distance according to the first formula, the first formula is: d represents the distance between the tree and the power line, v represents the direction vector of the power line, p 0 represents the coordinates of any coordinate point on the above power line, and q represents the coordinates of any tree in the voxelized area;

[0076] Step S2052, comparing the above distance with the above first tree barrier safety distance;

[0077] Step S2053, when the distance is less than or equal to the first tree barrier safety distance, determining the division result as the tree barrier area;

[0078] Step S2054: when the distance is greater than the first tree barrier safety distance, determining the division result as the non-tree barrier area.

[0079] In the above embodiment, the distance d between the power line and the tree voxel grid is calculated by using the classified power lines and voxelized trees, and the calculation result is divided to exclude the non-tree barrier area obtained in the first step of rapid detection. Specifically, assuming that the power line vector v represents the direction vector of the power line, the distance d is calculated as follows: According to the calculation results, the distance d between the power line and the tree voxel grid is gradient divided for the elimination of suspected tree barrier hazards, that is, the above distance is compared with the above first tree barrier safety distance. When the above distance is less than or equal to the above first tree barrier safety distance, the above division result is the above tree barrier area; when the above distance is greater than the above first tree barrier safety distance, the above division result is the above non-tree barrier area. The non-tree barrier area can be directly excluded and will no longer participate in the subsequent judgment of the tree barrier hazard level.

[0080] In order to improve the efficiency and accuracy of tree barrier assessment, in an optional implementation manner, the above step S206 includes:

[0081] Step S2061, determining the minimum tree obstacle safety distance between the tree and the transmission line according to the power of the transmission line;

[0082] Step S2062, when the distance is less than or equal to the first tree barrier safety distance and greater than the second tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the low hazard level, and the second tree barrier safety distance is less than the first tree barrier safety distance and greater than the minimum tree barrier safety distance;

[0083] Step S2063, when the distance is less than or equal to the second tree barrier safety distance and greater than the minimum tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the medium hazard level;

[0084] Step S2064: when the distance is less than or equal to the minimum tree obstacle safety distance, determine that the tree obstacle hazard level of the tree obstacle area corresponding to the target suspected tree obstacle position is the high hazard level.

[0085] In the above embodiment, the power of the transmission line determines its ability to transmit electric energy and the magnitude of the current. The safety distance requirements for tree barriers for higher power lines are usually more stringent. If the detected tree barrier is in the protection zone of the transmission line and the power of the line is high, it is necessary to determine that the tree barrier is of a higher hidden danger level, otherwise it is of a lower hidden danger level or a medium hidden danger level. According to the evaluation results, the hidden dangers of tree barriers are divided into three different levels, namely: low risk, medium risk, and high risk. These levels are comprehensively judged based on factors such as the distance between the tree barrier and the line and the power level of the line. The minimum tree barrier safety distance corresponding to the power of the transmission line can be determined in combination with the "Overhead Transmission Line Operation Regulations", and the different distances between the trees around the transmission lines of different powers and the transmission lines are interpreted to determine the tree barrier hidden danger level, and the detection results are output and displayed. By calculating the distance between the trees and the power lines and automatically evaluating the hidden danger level in combination with the power of the transmission lines, the evaluation efficiency and accuracy are greatly improved, the uncertainty caused by human judgment is reduced, and the intelligence level of power line inspection is improved.

[0086] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0087] The embodiment of the present application also provides a device for automatically identifying tree barrier hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles. It should be noted that the device for automatically identifying tree barrier hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles in the embodiment of the present application can be used to execute the method for automatically identifying tree barrier hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and those that have been explained will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation in hardware, or a combination of software and hardware, is also possible and conceived.

[0088] The following is an introduction to the automatic identification device for tree obstacle hazards in power transmission lines based on drone multi-source remote sensing provided in an embodiment of the present application.

[0089] Figure 5 1 is a structural block diagram of a device for automatically identifying tree obstacles in power transmission lines based on multi-source remote sensing of drones according to an embodiment of the present application. Figure 5 The device comprises:

[0090] The acquisition unit 10 is used to acquire multi-source remote sensing data, the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images are remote sensing images of power lines of transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers.

[0091] Specifically, when planning the route of a drone, it is necessary to comprehensively consider the direction and location of power lines to ensure the safety and efficiency of the route. Multi-source remote sensing refers to a technology that uses remote sensing data from different sensors, different time periods or different spatial resolutions for comprehensive analysis and processing. In the drone platform, it includes high-resolution images and LiDAR point cloud data. The drone route is planned according to the direction and location of the power lines, and the drone multi-sensor platform is used to obtain multi-source remote sensing data, including high-resolution remote sensing images and airborne LiDAR point cloud data.

[0092] The feature processing unit 20 is used to use the above-mentioned high-resolution remote sensing image to perform image feature extraction and feature fusion processing based on the improved target detection model to detect all suspected tree obstacle locations. The above-mentioned improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of the YOLOv8 model.

[0093] Specifically, the high-resolution remote sensing images are used to extract and fuse image features based on the improved YOLO v8 model to detect and obtain the suspected tree obstacle locations. In the feature fusion process of the YOLO v8 model, the SENet channel attention mechanism is introduced to improve the expression ability of tree obstacle features by adaptively adjusting the weight of each channel, reduce the impact of low-level vegetation on detection, and thus improve the model detection performance.

[0094] The extraction unit 30 is used to classify the above-mentioned airborne point cloud data based on the improved point cloud deep learning model to extract the power line results. The above-mentioned improved point cloud deep learning model is an improved model obtained by adding a special classifier to the output layer of the PointNet model. The above-mentioned special classifier is a classifier dedicated to separating linear point cloud data.

[0095] Specifically, on the basis of the original PointNet model, a special classifier is added to the output layer, LineSegNet (Line Segment Network), which is a classifier dedicated to separating linear point cloud structures. It is called the PointNet-PL model (i.e., the above-mentioned improved point cloud deep learning model). This model can further optimize the network's ability to separate linear point clouds, and can simultaneously combine geometric features and contextual information to finely classify power lines in point clouds; in addition, a multi-task learning framework is adopted to enhance the classifier's attention to power lines through a weighted loss function to ensure the accuracy of the network in identifying power lines. Through this improvement, the above-mentioned airborne point cloud data is input into the above-mentioned improved point cloud deep learning model for classification, and the power line results are extracted, which can more accurately extract power line features and effectively improve the high-precision classification of power line point clouds.

[0096] The retrieval classification unit 40 is used to perform the retrieval classification step, retrieve the above-mentioned airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the above-mentioned target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize and classify the above-mentioned target point cloud data to obtain a tree classification result. The above-mentioned target suspected tree obstacle position is any one of the above-mentioned suspected tree obstacle positions. The above-mentioned point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model.

[0097] Specifically, the suspected tree obstacle detection frame obtained by the YOLO v8 network is used to map the position of the detection frame in the image coordinate system to the three-dimensional space coordinate system, and a corresponding three-dimensional retrieval area is defined to filter out the point cloud falling in the area from the original point cloud data, that is, the above-mentioned target point cloud data. In the VoxelNet model, the GELU activation function is used to replace the traditional ReLU function to increase the feature expression ability of the network structure, so that it can obtain accurate tree voxels based on the tree structure characteristics more delicately, and it is called the VoxelNet-TR model (that is, the above-mentioned improved point cloud object detection model). The above-mentioned target point cloud data is input into the VoxelNet-TR model for voxel classification to obtain the tree classification result.

[0098] It should be noted that the present invention does not calculate all point cloud data, which reduces the amount of calculation and improves model efficiency. The PointNet-PL model (i.e., the above-mentioned improved point cloud deep learning model) is used for power line extraction, and the VoxelNet-TR model (i.e., the above-mentioned improved point cloud object detection model) performs voxel classification on the tree barrier point cloud, achieving accurate separation and identification of tree barriers and power lines, avoiding the calculation of all point cloud data, reducing the model's dependence on computer performance, and improving calculation efficiency.

[0099] The calculation unit 50 is used to execute the calculation step, calculate the distance between the trees and the power lines within the coverage area according to the tree classification results and the power line results, and perform gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance.

[0100] Specifically, the distance between the power lines and the tree voxel grids is calculated using the classified power lines and voxelized trees, and the calculation results are divided to obtain the division results so as to exclude the non-tree barrier areas.

[0101] The determination unit 60 is used to execute the determination step. When the above-mentioned division result is the above-mentioned tree barrier area, an analysis is performed based on the corresponding above-mentioned distance combined with the power of the above-mentioned transmission line to determine the tree barrier hazard level of the above-mentioned tree barrier area where the above-mentioned target suspected tree barrier position is located. The above-mentioned tree barrier hazard level is one of the following: high hazard level, medium hazard level and low hazard level.

[0102] Specifically, based on the determination of the tree barrier area, combined with the power level of the transmission line, the tree barrier hazards and the potential impact of tree barriers on the operation of the transmission line are further evaluated. The power of the transmission line determines its ability to transmit electrical energy and the magnitude of the current. Lines with higher power usually have stricter safety distance requirements for tree barriers. For tree barriers within the line protection area, the tree barrier hazard level is determined based on the distance outside the edge line determined in the "Overhead Transmission Line Operation Regulations".

[0103] The repetition unit 70 is used to repeatedly execute the above-mentioned retrieval classification step, the above-mentioned calculation step and the above-mentioned determination step in sequence at least once until the judgment of all the above-mentioned suspected tree barrier positions is completed, and all the above-mentioned tree barrier areas and the corresponding above-mentioned tree barrier hazard levels are obtained, so as to inspect the corresponding above-mentioned tree barrier areas according to the above-mentioned tree barrier hazard levels.

[0104] Specifically, the above-mentioned retrieval and classification step, the above-mentioned calculation step and the above-mentioned determination step are iteratively executed until the interpretation of all suspected tree barrier locations is completed, all non-tree barrier areas are gradually excluded, and the real tree barrier areas and the corresponding hidden danger levels are graded and displayed, so that the corresponding tree barrier areas can be inspected according to the above-mentioned tree barrier hidden danger levels to ensure the accuracy and comprehensiveness of the inspection results.

[0105] In this embodiment, the acquisition unit is used to acquire multi-source remote sensing data, the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images are remote sensing images of power lines of transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers; the feature processing unit is used to use the high-resolution remote sensing images to perform image feature extraction and feature fusion processing based on the improved target detection model to detect all suspected tree obstacle positions, the improved target detection model is an improved model obtained by introducing the channel attention mechanism on the basis of the YOLOv8 model; the extraction unit is used to extract and fuse the image features based on the improved target detection model. The point cloud deep learning model is used to classify the above-mentioned airborne point cloud data and extract the power line results. The above-mentioned improved point cloud deep learning model is an improved model obtained by adding a special classifier on the output layer of the PointNet model. The above-mentioned special classifier is a classifier dedicated to separating linear point cloud data; a retrieval classification unit is used to perform the retrieval classification step, retrieve the above-mentioned airborne point cloud data within the coverage range where the target suspected tree obstacle position is located, obtain the target point cloud data corresponding to the above-mentioned target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize and classify the above-mentioned target point cloud data to obtain the tree classification result. The above-mentioned target suspected tree obstacle position is any The above-mentioned suspected tree barrier position is intended, and the above-mentioned point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model; a calculation unit is used to execute a calculation step, calculate the distance between the trees and the power lines within the above-mentioned coverage range according to the above-mentioned tree classification results and the above-mentioned power line results, and perform gradient division according to the above-mentioned distance to obtain a division result, and the above-mentioned division result is a tree barrier area or a non-tree barrier area, the above-mentioned tree barrier area is an area where the above-mentioned distance is less than the first tree barrier safety distance, and the above-mentioned non-tree barrier area is an area where the above-mentioned distance is greater than the above-mentioned first tree barrier safety distance; a determination unit, Used to execute the determination step, when the above division result is the above tree barrier area, analyze according to the corresponding above distance combined with the power of the above transmission line to determine the tree barrier hazard level of the above tree barrier area where the above target suspected tree barrier position is located, and the above tree barrier hazard level is one of the following: high hazard level, medium hazard level and low hazard level; a repeating unit, used to repeat the above retrieval classification step, the above calculation step and the above determination step in sequence at least once until all the above suspected tree barrier positions are judged, and all the above tree barrier areas and the corresponding tree barrier hazard levels are obtained, so as to inspect the corresponding tree barrier areas according to the above tree barrier hazard levels. This application gives full play to the advantages of image information and point cloud data collected by drones, and processes them directly.Based on the improved YOLO v8 model, the image features of the acquired high-resolution remote sensing images are extracted and fused to detect and obtain the location of suspected tree barriers in order to further analyze the potential risk areas; then, the improved PointNet model is used to classify the point cloud data, extract the transmission lines, and retrieve the point cloud data in the area according to the detected suspected tree barrier locations. Subsequently, the improved VoxelNet model is used to classify these point cloud data to identify and voxelize the tree areas to provide support for subsequent precise analysis; finally, the distance between the trees and the wires is calculated based on the classification results, and the distance calculation results are divided to exclude non-tree barrier areas, thereby achieving accurate interpretation of the area. Combined with the power of the transmission line, the tree barrier hazard level is determined, and the final detection results are output and displayed. Compared with the existing methods, the method of the present application can obtain the tree barrier extraction results more quickly and accurately. The present application solves the problem of low efficiency in the inspection of tree barriers on transmission lines in the prior art.

[0106] In order to improve the accuracy and comprehensiveness of tree obstacle detection, in an optional implementation manner, the acquisition unit includes:

[0107] A planning module plans the flight trajectory of the UAV according to the direction and location distribution of the above-mentioned power lines to obtain the flight trajectory of the target UAV;

[0108] The acquisition module controls the medium and high resolution remote sensing equipment carried by the UAV platform to obtain the fine image data of the power lines according to the flight trajectory of the target UAV. At the same time, the laser radar carried by the above UAV platform is used to obtain the above airborne point cloud data.

[0109] In the above embodiments, the flight trajectory of the UAV is manually planned according to the direction and location distribution of the transmission line to ensure that the flight path can fully cover the key areas of the transmission line and its surrounding environment; by setting the appropriate flight altitude, speed and route of the UAV, the high-resolution remote sensing equipment carried by the UAV platform is used to obtain fine image data of the power line, accurately capture the details of the line and the changes in the surrounding environment. The multispectral sensor equipment obtains high-resolution image data to capture the details of the transmission line and the changes in the surrounding environment; at the same time, the laser radar system carried by the UAV platform is used to obtain three-dimensional LiDAR point cloud data to achieve spatial positioning of different targets in the transmission line (such as power lines, trees, transmission towers, etc.), which can effectively cover the transmission line in different environments, and improve the accuracy and comprehensiveness of the detection. In this example, high-resolution remote sensing images are obtained by using optical sensors for ground observation and collection; airborne laser radar point cloud data are obtained by using airborne laser radar sensors for ground observation and collection.

[0110] In order to improve the expression ability of tree obstacle features to accurately obtain the suspected tree obstacle location, in an optional implementation, the feature processing unit includes:

[0111] The first preprocessing module sequentially performs image resizing, normalization and image enhancement on the high-resolution remote sensing image to obtain a remote sensing image of target resolution;

[0112] The first input module inputs the target resolution remote sensing image into the improved target detection model, and uses the Backbone module in the improved target detection model to extract the image features of the target resolution remote sensing image through multi-layer convolution operations to generate the multi-scale feature map, uses the Neck module to fuse the multi-scale feature map through a feature pyramid network and a path aggregation network to obtain a fused feature map, uses the Head module to detect the fused feature map and outputs all the suspected tree obstacle positions in sequence to obtain all the suspected tree obstacle positions.

[0113] In the above embodiment, data preprocessing is first performed to normalize and enhance the high-resolution remote sensing images to improve the robustness and generalization ability of the model. Figure 3As shown in the upper part, the target high-resolution remote sensing image is input into the improved YOLO v8 model. The model structure of YOLO v8 mainly includes three parts: Backbone, Neck and Head. Each part plays a key role in feature extraction, fusion and target detection. The Backbone module is used to extract image features and generate multi-scale feature maps. The Neck module is used to perform feature fusion processing on the multi-scale feature maps to generate fused feature maps. The Head module is used to detect the fused feature maps and output them. The preprocessed image (the target resolution remote sensing image) is input into the Backbone module for feature extraction. The core of this stage is to extract the low-level to high-level feature maps of the image through multi-layer convolution operations. Backbone uses a deep neural network to ensure that key visual information can be fully expressed. The Backbone module mainly extracts individual scale features from the images generated by the C2f module and the SPPF module for feature extraction. The C2f module replaces the original C3 module. After reducing one convolution layer, the model is more lightweight. SPPF reduces the number of network layers on the basis of the original SPP, which can perform feature fusion faster and eliminate redundant operations. Next, the extracted multi-scale feature maps are passed to the Neck module. The Neck module can effectively fuse feature maps of different resolutions and semantic levels through the combination of feature pyramid network (FPN) and path aggregation network (PAN). FPN enhances the semantic expression ability of higher-level feature maps through top-down feature fusion, ensuring that the model can understand complex objects in the image. At the same time, PAN retains the detailed information of lower-level feature maps through bottom-up path enhancement, making the model more accurate when detecting targets of different sizes. Through the introduction of SENet channel attention mechanism, the weight of each channel can be adaptively adjusted, and the global information of each channel is extracted by global average pooling. Then, weights are generated through two layers of fully connected networks. After the weight distribution between channels, the network can focus on more meaningful channel features and suppress irrelevant or redundant information. This mechanism enhances the model's ability to capture important features and improves the accuracy of target detection, especially when dealing with complex scenes and multi-scale targets. After feature fusion is completed, the output of the Neck module is passed to the Head module, which is responsible for converting the fused feature map into a specific detection result. Through a series of convolution operations and classifiers, the model can output the specific location and category information of the tree barrier. In the feature fusion process of the YOLO v8 model, the SENet channel attention mechanism is introduced to improve the expression ability of tree barrier features by adaptively adjusting the weight of each channel, reduce the impact of low-level vegetation on detection, and thus improve the model detection performance.

[0114] In order to more accurately extract the power line features and effectively improve the high-precision classification of the power line point cloud, in an optional implementation, the extraction unit includes:

[0115] A second preprocessing module performs a removal operation and a normalization process on the airborne point cloud data in sequence to obtain valid airborne point cloud data, wherein the removal operation is an operation of removing point cloud data that is not related to the power line;

[0116] The second input module inputs the above-mentioned valid airborne point cloud data into the above-mentioned improved point cloud deep learning model, so that the above-mentioned transformation network in the above-mentioned improved point cloud deep learning model aligns the above-mentioned valid airborne point cloud data to the standard transformation matrix through multiple convolutional layers and fully connected layers, the above-mentioned feature extraction network uses a multi-layer perceptron to extract local point features from the above-mentioned transformation matrix, the maximum pooling layer in the above-mentioned global feature aggregation and classification module aggregates the above-mentioned local point features into a global feature vector, and the above-mentioned global feature vector is remapped through the above-mentioned special classifier in the above-mentioned output layer to generate linear point cloud features for classification output, so as to obtain the above-mentioned power line result.

[0117] In the above embodiment, the network structure of the above improved point cloud deep learning model includes a transformation network, a feature extraction network and a global feature aggregation and classification output module, and the above global feature aggregation and classification output module includes at least a maximum pooling layer and the above output layer. Remove irrelevant point cloud data such as ground, building and other noise data from the collected airborne point cloud data to reduce the impact of errors on classification, and normalize the point cloud data so that the spatial positions of all points are standardized to a unified coordinate range to obtain the above valid airborne point cloud data. Figure 4 As shown in the figure, first, the point cloud data (that is, the valid airborne point cloud data) is input into the transformation network using the T-Net structure. Through multiple 3×3 convolutional layers and fully connected layers, the transformation matrix that aligns the input valid airborne point cloud data to the standard posture is learned. The feature extraction network uses point-by-point MLP (multi-layer perceptron) to extract point features, where each layer of MLP is equivalent to a 1×1 convolutional layer to extract the local features of the point, and the activation function is ReLU. In this process, no spatial pooling operation is performed to maintain the spatial information of the point. The feature vector output by the feature extraction network is aggregated into a global feature vector through the maximum pooling operation. This operation integrates the local information of the point cloud into a global descriptor of a fixed length, which is invariant to spatial transformation. Finally, through several fully connected layers, the network maps the global features to the output space of a specific task (such as classification or segmentation), and finally calculates the category probability through the Softmax layer or realizes point cloud segmentation through the point-by-point MLP.

[0118] In this embodiment, if Figure 4As shown in the figure, the PointNet model is improved as follows: Based on the original PointNet model, a special classifier is added to the output layer, LineSegNet (Line Segment Network) is a classifier dedicated to separating linear point cloud structures; in addition, a multi-task learning framework is adopted to enhance the classifier's attention to power lines through a weighted loss function to ensure the accuracy of the network in identifying power lines. The improved model is called the PointNet-PL model (i.e. the above-mentioned improved point cloud deep learning model). Based on the PointNet-PL model, the power lines are extracted by classifying the point cloud data. The LineSegNet classifier is added to the feature aggregation module of the original PointNet. The classifier remaps the aggregated global features through a dedicated MLP layer to generate specific features of the linear point cloud. In this process, the structure of the MLP layer is consistent with the original PointNet, but an additional fully connected layer with specific weight initialization is added before the output layer to enhance the recognition of linear structures. In the output layer of the classifier, a weighted loss function is introduced to give a higher weight specifically to the power line category, thereby enhancing the classifier's attention to the power line. Through this improvement, the PointNet-PL model can extract power line features more accurately and effectively improve the high-precision classification of power line point clouds.

[0119] In order to ensure that tree information can be detected more accurately in complex scenes, in an optional implementation, the retrieval classification unit includes:

[0120] A retrieval module, using the improved target detection model to retrieve a suspected tree obstacle detection frame corresponding to the suspected tree obstacle position of the target;

[0121] A mapping module maps the coordinate position of the suspected tree obstacle detection frame in the image coordinate system to the three-dimensional space coordinate system, and sets a corresponding three-dimensional search area;

[0122] A screening module is used to screen out the airborne point cloud data falling within the three-dimensional search area to obtain the target point cloud data;

[0123] The third input module inputs the target point cloud data into the improved point cloud object detection model, so that the voxelization module in the improved point cloud object detection model divides the target point cloud data into multiple voxel grids of set sizes, performs local feature extraction on the target point cloud data in the non-empty voxel grids through the voxel feature encoding module to obtain voxel features, and further extracts spatial features between voxels for all the voxel features through the convolutional intermediate layer, and the region proposal network classifies and locates trees in the three-dimensional retrieval area according to the spatial features to obtain the tree classification result.

[0124] In the above embodiment, the network structure of the improved point cloud object detection model includes a voxelization module, a feature extraction module and a classification regression module. The voxelization module is used to perform voxelization operations, the feature extraction module includes a voxel feature encoding module and a convolutional intermediate layer, and the classification regression module is used to classify and locate the target object through a region proposal network. First, the suspected tree obstacle detection frame retrieved by the improved YOLO v8 model is mapped to the position of the suspected tree obstacle detection frame in the image coordinate system into the three-dimensional space coordinate system, and a corresponding three-dimensional retrieval area is defined to filter out the point cloud (i.e., the target point cloud data) falling within the area from the original point cloud data. The target point cloud data is divided into a three-dimensional voxel grid of a fixed size by the improved VoxelNet model. The point cloud in each non-empty voxel is processed by the VFE module to extract local features. The VFE module uses a multi-layer perceptron (MLP) with shared parameters to perform feature mapping on the coordinates of each point, cancels the traditional ReLU function, and applies the GELU activation function. This process can be expressed as: ij =GELU(W 1 p ij +b 1 ), where p ij Represents voxel v j The coordinates of the i-th point in W 1 and b 1are the weight matrix and bias term of MLP respectively. The network is downsampled through the maximum pooling layer after the convolution operation, and the spatial size of the feature map is reduced layer by layer. The second and third layers use 128 and 256 convolution kernels for feature extraction, respectively. The spatial size of the final feature map is reduced to 1 / 8 of the original voxel grid, but the number of channels is increased to 256. On the feature map extracted by 3D convolution, the region proposal network (RPN) classifies and locates the target object. RPN contains a 3D convolution layer, which is then divided into a classification branch and a regression branch. The former calculates the category probability through Softmax, and the latter outputs the bounding box regression parameters. The entire network structure uses the GELU activation function to improve the nonlinear expression ability, so that the improved VoxelNet model (also known as the VoxelNet-TR model) can more accurately detect tree information in complex scenes.

[0125] In order to accurately exclude non-tree barrier areas, in an optional implementation manner, the calculation unit includes:

[0126] The calculation module calculates the distance according to the first formula. The first formula is: d represents the distance between the tree and the power line, v represents the direction vector of the power line, p0 represents the coordinates of any coordinate point on the power line, and q represents the coordinates of any tree in the voxelized area;

[0127] A comparison module compares the distance with the first tree barrier safety distance;

[0128] A first determining module, when the distance is less than or equal to the first tree barrier safety distance, determines the division result as the tree barrier area;

[0129] The second determining module determines that the division result is the non-tree barrier area when the distance is greater than the first tree barrier safety distance.

[0130] In the above embodiment, the distance d between the power line and the tree voxel grid is calculated by using the classified power lines and voxelized trees, and the calculation result is divided to exclude the non-tree barrier area obtained in the first step of rapid detection. Specifically, assuming that the power line vector v represents the direction vector of the power line, the distance d is calculated as follows: According to the calculation results, the distance d between the power line and the tree voxel grid is gradient divided for the elimination of suspected tree barrier hazards, that is, the above distance is compared with the above first tree barrier safety distance. When the above distance is less than or equal to the above first tree barrier safety distance, the above division result is the above tree barrier area; when the above distance is greater than the above first tree barrier safety distance, the above division result is the above non-tree barrier area. The non-tree barrier area can be directly excluded and will no longer participate in the subsequent judgment of the tree barrier hazard level.

[0131] In order to improve the efficiency and accuracy of tree barrier assessment, in an optional implementation manner, the determination unit includes:

[0132] A third determination module determines a minimum tree obstacle safety distance between the tree and the transmission line according to the power of the transmission line;

[0133] A fourth determination module determines, when the distance is less than or equal to the first tree barrier safety distance and greater than the second tree barrier safety distance, that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the low hazard level, and the second tree barrier safety distance is less than the first tree barrier safety distance and greater than the minimum tree barrier safety distance;

[0134] A fifth determination module, when the distance is less than or equal to the second tree barrier safety distance and greater than the minimum tree barrier safety distance, determines that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the medium hazard level;

[0135] The sixth determination module determines that the tree obstacle hazard level of the tree obstacle area corresponding to the target suspected tree obstacle position is the high hazard level when the distance is less than or equal to the minimum tree obstacle safety distance.

[0136] In the above embodiment, the power of the transmission line determines its ability to transmit electric energy and the magnitude of the current. The safety distance requirements for tree barriers for higher power lines are usually more stringent. If the detected tree barrier is in the protection zone of the transmission line and the power of the line is high, it is necessary to determine that the tree barrier is of a higher hidden danger level, otherwise it is of a lower hidden danger level or a medium hidden danger level. According to the evaluation results, the hidden dangers of tree barriers are divided into three different levels, namely: low risk, medium risk, and high risk. These levels are comprehensively judged based on factors such as the distance between the tree barrier and the line and the power level of the line. The minimum tree barrier safety distance corresponding to the power of the transmission line can be determined in combination with the "Overhead Transmission Line Operation Regulations", and the different distances between the trees around the transmission lines of different powers and the transmission lines are interpreted to determine the tree barrier hidden danger level, and the detection results are output and displayed. By calculating the distance between the trees and the power lines and automatically evaluating the hidden danger level in combination with the power of the transmission lines, the evaluation efficiency and accuracy are greatly improved, the uncertainty caused by human judgment is reduced, and the intelligence level of power line inspection is improved. The above-mentioned automatic identification device for tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles includes a processor and a memory. The above-mentioned acquisition unit, feature processing unit and extraction unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in the form of any combination.

[0137] The processor includes a kernel, and the kernel calls the corresponding program unit from the memory. One or more kernels can be set, and the problem of low inspection efficiency of tree barriers on power transmission lines in the prior art can be solved by adjusting kernel parameters.

[0138] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0139] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the method for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles.

[0140] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program, when running, executes the method for automatically identifying tree obstacle hazards in power transmission lines based on multi-source remote sensing of unmanned aerial vehicles.

[0141] An embodiment of the present invention provides a system for automatically identifying tree obstacles and hidden dangers in power transmission lines. The system includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, at least the following steps are implemented:

[0142] Step S201, acquiring multi-source remote sensing data, the multi-source remote sensing data at least including high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images are remote sensing images of power lines of transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers;

[0143] Step S202, using the high-resolution remote sensing image to perform image feature extraction and feature fusion processing based on an improved target detection model to detect all suspected tree obstacle locations, wherein the improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of a YOLOv8 model;

[0144] Step S203, classifying the above-mentioned airborne point cloud data based on the improved point cloud deep learning model to extract the power line results, wherein the above-mentioned improved point cloud deep learning model is an improved model obtained by adding a special classifier to the output layer of the PointNet model, and the above-mentioned special classifier is a classifier dedicated to separating linear point cloud data;

[0145] Step S204, a retrieval and classification step, retrieving the above-mentioned airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtaining the target point cloud data corresponding to the above-mentioned target suspected tree obstacle position, and performing voxel classification on the above-mentioned target point cloud data using the improved point cloud object detection model to obtain a tree classification result, wherein the above-mentioned target suspected tree obstacle position is any one of the above-mentioned suspected tree obstacle positions, and the above-mentioned point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model;

[0146] Step S205, a calculation step, according to the tree classification result and the power line result, calculating the distance between the trees and the power lines within the coverage area, and performing gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance;

[0147] Step S206, a determination step, in the case where the division result is the tree barrier area, an analysis is performed based on the corresponding distance combined with the power of the transmission line to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, and the tree barrier hazard level is one of the following: a high hazard level, a medium hazard level, and a low hazard level;

[0148] Step S207, repeatedly executing the above-mentioned retrieval classification step, the above-mentioned calculation step and the above-mentioned determination step at least once in sequence until the judgment of all the above-mentioned suspected tree barrier positions is completed, and all the above-mentioned tree barrier areas and the corresponding above-mentioned tree barrier hazard levels are obtained, so as to inspect the corresponding above-mentioned tree barrier areas according to the above-mentioned tree barrier hazard levels.

[0149] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0150] Step S201, acquiring multi-source remote sensing data, the multi-source remote sensing data at least including high-resolution remote sensing images and airborne point cloud data, the high-resolution remote sensing images are remote sensing images of power lines of transmission lines, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees and transmission towers;

[0151] Step S202, using the high-resolution remote sensing image to perform image feature extraction and feature fusion processing based on an improved target detection model to detect all suspected tree obstacle locations, wherein the improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of a YOLOv8 model;

[0152] Step S203, classifying the above-mentioned airborne point cloud data based on the improved point cloud deep learning model to extract the power line results, wherein the above-mentioned improved point cloud deep learning model is an improved model obtained by adding a special classifier to the output layer of the PointNet model, and the above-mentioned special classifier is a classifier dedicated to separating linear point cloud data;

[0153] Step S204, a retrieval and classification step, retrieving the above-mentioned airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtaining the target point cloud data corresponding to the above-mentioned target suspected tree obstacle position, and performing voxel classification on the above-mentioned target point cloud data using the improved point cloud object detection model to obtain a tree classification result, wherein the above-mentioned target suspected tree obstacle position is any one of the above-mentioned suspected tree obstacle positions, and the above-mentioned point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as the activation function based on the VoxelNet model;

[0154] Step S205, a calculation step, according to the tree classification result and the power line result, calculating the distance between the trees and the power lines within the coverage area, and performing gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than the first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance;

[0155] Step S206, a determination step, in the case where the division result is the tree barrier area, an analysis is performed based on the corresponding distance combined with the power of the transmission line to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, and the tree barrier hazard level is one of the following: a high hazard level, a medium hazard level, and a low hazard level;

[0156] Step S207, repeatedly executing the above-mentioned retrieval classification step, the above-mentioned calculation step and the above-mentioned determination step at least once in sequence until the judgment of all the above-mentioned suspected tree barrier positions is completed, and all the above-mentioned tree barrier areas and the corresponding above-mentioned tree barrier hazard levels are obtained, so as to inspect the corresponding above-mentioned tree barrier areas according to the above-mentioned tree barrier hazard levels.

[0157] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0158] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0159] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0162] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0163] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0164] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. 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 technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0165] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0166] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0167] 1) The automatic identification method of tree barrier hidden dangers of power transmission lines based on multi-source remote sensing of unmanned aerial vehicles in this application fully utilizes the advantages of image information and point cloud data collected by unmanned aerial vehicles and directly processes them. Based on the improved YOLO v8 model, the image features of the acquired high-resolution remote sensing images are extracted and fused, and the location of suspected tree barriers is detected and obtained, so as to further analyze the potential risk areas; then, the point cloud data is classified using the improved PointNet model, the transmission lines are extracted, and the point cloud data in the area is retrieved according to the detected suspected tree barrier location. Subsequently, the point cloud data are classified using the improved VoxelNet model to identify and voxelize the tree area to provide support for subsequent precise analysis; finally, the distance between the tree and the wire is calculated according to the classification results, and the distance calculation results are divided to exclude the non-tree barrier area, so as to achieve accurate interpretation of the area. Combined with the power of the transmission line, the tree barrier hidden danger level is determined, and the final detection results are output and displayed. Compared with the existing methods, the method of this application can obtain the tree barrier extraction results more quickly and accurately. This application solves the problem of low inspection efficiency of tree barriers on power transmission lines in the prior art.

[0168] 2) The automatic identification device for tree barrier hidden dangers of power transmission lines based on multi-source remote sensing of unmanned aerial vehicles of the present application gives full play to the advantages of image information and point cloud data collected by unmanned aerial vehicles, and directly processes them. Based on the improved YOLO v8 model, the image features of the acquired high-resolution remote sensing images are extracted and fused, and the location of suspected tree barriers is detected and obtained, so as to further analyze the potential risk areas; then, the point cloud data is classified by using the improved PointNet model, the transmission lines are extracted, and the point cloud data in the area is retrieved according to the detected suspected tree barrier location. Subsequently, the point cloud data are classified using the improved VoxelNet model, so as to identify and voxelize the tree area, providing support for subsequent precise analysis; finally, the distance between the tree and the wire is calculated according to the classification result, and the distance calculation result is divided to exclude the non-tree barrier area, so as to achieve accurate interpretation of the area. Combined with the power of the transmission line, the tree barrier hidden danger level is determined, and the final detection result is output and displayed. Compared with the existing method, the method of the present application can obtain the tree barrier extraction result more quickly and accurately. The present application solves the problem of low inspection efficiency of tree barriers of power transmission lines in the prior art.

[0169] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for automatically identifying tree obstacles in power transmission lines based on multi-source remote sensing by unmanned aerial vehicles, characterized in that: include: Acquire multi-source remote sensing data, wherein the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, wherein the high-resolution remote sensing images are remote sensing images of power lines of the transmission line, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees, and transmission towers; The high-resolution remote sensing image is used to perform image feature extraction and feature fusion processing based on an improved target detection model to detect all suspected tree obstacle locations. The improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of the YOLOv8 model. Classifying the airborne point cloud data based on an improved point cloud deep learning model to extract power line results, wherein the improved point cloud deep learning model is an improved model obtained by adding a dedicated classifier to the output layer of the PointNet model, and the dedicated classifier is a classifier dedicated to separating linear point cloud data; A retrieval and classification step, retrieving the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtaining the target point cloud data corresponding to the target suspected tree obstacle position, and performing voxel classification on the target point cloud data using an improved point cloud object detection model to obtain a tree classification result, wherein the target suspected tree obstacle position is any one of the suspected tree obstacle positions, and the point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as an activation function based on the VoxelNet model; a calculation step, calculating the distance between the trees and the power lines within the coverage area according to the tree classification result and the power line result, and performing gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than a first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance; A determination step, when the division result is the tree barrier area, analyzing according to the corresponding distance combined with the power of the transmission line, to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, wherein the tree barrier hazard level is one of the following: a high hazard level, a medium hazard level, and a low hazard level; The retrieval and classification step, the calculation step and the determination step are sequentially repeated at least once until all the suspected tree barrier locations are judged, and all the tree barrier areas and the corresponding tree barrier hazard levels are obtained, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels.

2. The method according to claim 1, characterized in that Access multi-source remote sensing data, including: Planning the flight trajectory of the UAV according to the direction and position distribution of the power lines to obtain the flight trajectory of the target UAV; According to the flight trajectory of the target UAV, the medium and high resolution remote sensing equipment carried by the UAV platform is controlled to obtain the fine image data of the power lines. At the same time, the laser radar carried by the UAV platform is used to obtain the airborne point cloud data.

3. The method according to claim 1, characterized in that The improved target detection model includes a Backbone module, a Neck module and a Head module. The Backbone module is used to extract image features to generate a multi-scale feature map. The Neck module is used to perform feature fusion processing on the multi-scale feature map to generate a fused feature map. The Head module is used to detect and output the fused feature map. The high-resolution remote sensing image is used to perform image feature extraction and feature fusion processing based on the improved target detection model to detect all suspected tree obstacle locations, including: The high-resolution remote sensing image is sequentially subjected to image resizing, normalization processing and image enhancement processing to obtain a remote sensing image of target resolution; The target resolution remote sensing image is input into the improved target detection model, and the Backbone module in the improved target detection model is used to extract the image features of the target resolution remote sensing image through multi-layer convolution operations to generate the multi-scale feature map, and the Neck module is used to fuse the multi-scale feature map through a feature pyramid network and a path aggregation network to obtain a fused feature map, and the Head module is used to detect the fused feature map and output all the suspected tree obstacle positions in sequence to obtain all the suspected tree obstacle positions.

4. The method according to claim 1, characterized in that: The network structure of the improved point cloud deep learning model includes a transformation network, a feature extraction network, and a global feature aggregation and classification output module, wherein the global feature aggregation and classification output module includes at least a maximum pooling layer and the output layer. The airborne point cloud data is classified based on the improved point cloud deep learning model to extract power line results, including: The airborne point cloud data is sequentially subjected to removal and normalization processing to obtain valid airborne point cloud data, wherein the removal operation is an operation of removing point cloud data irrelevant to the power line; The valid airborne point cloud data is input into the improved point cloud deep learning model, so that the transformation network in the improved point cloud deep learning model aligns the valid airborne point cloud data to a standard transformation matrix through multiple convolutional layers and fully connected layers, the feature extraction network uses a multi-layer perceptron to extract local point features from the transformation matrix, the maximum pooling layer in the global feature aggregation and classification module aggregates the local point features into a global feature vector, and the global feature vector is remapped through the dedicated classifier in the output layer to generate linear point cloud features for classification output, thereby obtaining the power line result.

5. The method according to claim 1, characterized in that The network structure of the improved point cloud object detection model includes a voxelization module, a feature extraction module and a classification and regression module. The voxelization module is used to perform voxelization operations. The feature extraction module includes a voxel feature encoding module and a convolutional intermediate layer. The classification and regression module is used to classify and locate the target object through a region proposal network, retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use the improved point cloud object detection model to voxelize and classify the target point cloud data to obtain a tree classification result, including: Retrieving a suspected tree obstacle detection frame corresponding to the suspected tree obstacle position of the target using the improved target detection model; Mapping the coordinate position of the suspected tree obstacle detection frame in the image coordinate system to the three-dimensional space coordinate system, and setting a corresponding three-dimensional search area; Filtering out the airborne point cloud data falling within the three-dimensional search area to obtain the target point cloud data; The target point cloud data is input into the improved point cloud object detection model, so that the voxelization module in the improved point cloud object detection model divides the target point cloud data into a plurality of voxel grids of set sizes, the voxel feature encoding module performs local feature extraction on the target point cloud data in the non-empty voxel grids to obtain voxel features, and the convolutional intermediate layer further extracts spatial features between voxels for all the voxel features, the region proposal network classifies and locates trees in the three-dimensional retrieval area according to the spatial features, and obtains the tree classification result.

6. The method according to claim 1, characterized in that According to the tree classification result and the power line result, the distance between the trees and the power lines within the coverage area is calculated, and gradient division is performed according to the distance to obtain a division result, including: The distance is calculated according to a first formula, wherein the first formula is: d represents the distance between the tree and the power line, v represents the direction vector of the power line, p0 represents the coordinates of any coordinate point on the power line, and q represents the coordinates of any tree in the voxelized area; Comparing the distance with the first tree barrier safety distance; When the distance is less than or equal to the first tree barrier safety distance, determining the division result as the tree barrier area; When the distance is greater than the first tree barrier safety distance, the division result is determined to be the non-tree barrier area.

7. The method according to claim 1, characterized in that In the case where the division result is the tree barrier area, performing analysis according to the corresponding distance in combination with the power of the transmission line to determine the tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located includes: Determining a minimum tree obstacle safety distance between the tree and the transmission line according to the power of the transmission line; When the distance is less than or equal to the first tree barrier safety distance and greater than the second tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the low hazard level, and the second tree barrier safety distance is less than the first tree barrier safety distance and greater than the minimum tree barrier safety distance; When the distance is less than or equal to the second tree barrier safety distance and greater than the minimum tree barrier safety distance, determining that the tree barrier hazard level of the tree barrier area corresponding to the target suspected tree barrier position is the medium hazard level; When the distance is less than or equal to the minimum tree obstacle safety distance, it is determined that the tree obstacle hazard level of the tree obstacle area corresponding to the target suspected tree obstacle position is the high hazard level.

8. An automatic identification device for tree obstacle hazards in power transmission lines based on multi-source remote sensing by unmanned aerial vehicles, characterized in that: The device comprises: An acquisition unit, used to acquire multi-source remote sensing data, wherein the multi-source remote sensing data at least includes high-resolution remote sensing images and airborne point cloud data, wherein the high-resolution remote sensing images are remote sensing images of power lines of the transmission line, and the airborne point cloud data at least includes spatial positioning data of the power lines, trees, and transmission towers; A feature processing unit is used to perform image feature extraction and feature fusion processing based on the improved target detection model using the high-resolution remote sensing image to detect all suspected tree obstacle locations, wherein the improved target detection model is an improved model obtained by introducing a channel attention mechanism on the basis of the YOLOv8 model; An extraction unit, used to classify the airborne point cloud data based on an improved point cloud deep learning model to extract power line results, wherein the improved point cloud deep learning model is an improved model obtained by adding a dedicated classifier to the output layer of the PointNet model, and the dedicated classifier is a classifier dedicated to separating linear point cloud data; A retrieval classification unit is used to perform a retrieval classification step, retrieve the airborne point cloud data within the coverage range of the target suspected tree obstacle position, obtain the target point cloud data corresponding to the target suspected tree obstacle position, and use an improved point cloud object detection model to voxelize and classify the target point cloud data to obtain a tree classification result, wherein the target suspected tree obstacle position is any one of the suspected tree obstacle positions, and the point cloud object detection model is an improved model obtained by replacing the RELU function with the GELU function as an activation function based on the VoxelNet model; a calculation unit, configured to execute a calculation step, calculate the distance between the trees and the power lines within the coverage area according to the tree classification result and the power line result, and perform gradient division according to the distance to obtain a division result, wherein the division result is a tree barrier area or a non-tree barrier area, wherein the tree barrier area is an area where the distance is less than a first tree barrier safety distance, and the non-tree barrier area is an area where the distance is greater than the first tree barrier safety distance; a determining unit, configured to execute a determining step, wherein, when the division result is the tree barrier area, an analysis is performed according to the corresponding distance in combination with the power of the transmission line to determine a tree barrier hazard level of the tree barrier area where the target suspected tree barrier position is located, wherein the tree barrier hazard level is one of the following: a high hazard level, a medium hazard level, and a low hazard level; A repeating unit is used to repeatedly execute the retrieval classification step, the calculation step and the determination step in sequence at least once until the judgment of all the suspected tree barrier positions is completed, and all the tree barrier areas and the corresponding tree barrier hazard levels are obtained, so as to inspect the corresponding tree barrier areas according to the tree barrier hazard levels.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. An automatic identification system for tree obstacles in power transmission lines, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of claims 1 to 7.

Images