Low-cost unmanned aerial vehicle autonomous inspection method and device for power overhead line

Abstract

The invention provides a low-cost unmanned aerial vehicle autonomous inspection method and device for an electric power overhead line. The method comprises the following steps: acquiring an inspection image of a to-be-inspected tower; identifying the type of the pole tower in the inspection image, and extracting the three-dimensional coordinate information of the pole tower; establishing a model library of various towers, and setting an unmanned aerial vehicle shooting point location and a flight path of each tower; generating a parameterized three-dimensional model of the identified tower in a three-dimensional space map based on the identified tower type, the three-dimensional coordinate information of the tower and a model library, and obtaining an unmanned aerial vehicle shooting point location and a flight path corresponding to the identified tower type; whether channel tree diseases exist in the inspection images or not is analyzed, and unmanned aerial vehicle shooting point locations and flight paths of towers with the channel tree diseases are optimized; generating a final unmanned aerial vehicle autonomous inspection track file; according to the invention, one-time investment construction and subsequent repeated application can be realized, the application cost of unmanned aerial vehicle autonomous inspection is greatly reduced, and low-cost distribution network unmanned aerial vehicle autonomous inspection is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous inspection by unmanned aerial vehicles (UAVs), and in particular to a low-cost autonomous inspection method and device for electric power overhead lines by unmanned aerial vehicles (UAVs). Background Art

[0002] In the field of power inspection, the traditional manual inspection mode is difficult to meet the needs of efficient operation and maintenance of power overhead lines because of its high reliance on manpower, time-consuming and limited efficiency. In recent years, the leap-forward progress of drone technology has brought revolutionary changes to this field. As an innovative inspection alternative, autonomous drone inspection is gradually emerging and showing strong application potential. The advantages of autonomous drone inspection are: first, using advanced 3D laser point cloud camera technology to perform high-precision 3D modeling of complex distribution network lines; then, based on these precise 3D coordinate information, a track planning file is generated to guide the drone to autonomously perform inspection tasks along the preset path, greatly improving the accuracy and efficiency of the inspection.

[0003] However, despite the huge technical advantages of autonomous drone inspections, their large-scale application faces severe challenges. Specifically, high-end 3D laser point cloud camera equipment is expensive, costing tens of thousands of yuan each, which constitutes an important threshold for initial investment; in addition, the laser point cloud modeling service fee is as high as one thousand yuan per kilometer. This cumulative cost effect undoubtedly adds a considerable economic burden to the inspection of large-scale overhead power lines. Therefore, the current high inspection cost has become a key factor restricting the in-depth application of drone technology in the field of overhead power line inspections, which in turn affects the improvement of overall inspection quality and efficiency.

[0004] At the same time, in complex mountainous conditions, overhead power lines usually pass through areas with a large number of trees. The growth of trees is uncontrollable, and over time, trees may invade power channels. Tree obstruction will bring greater safety hazards to autonomous drone inspections, and there is a risk of drone collisions.

[0005] In summary, although the prospect of autonomous inspection technology of UAV is broad, in order to realize large-scale application in power overhead line inspection, it is still necessary to continuously explore and optimize in terms of cost control and safety to overcome the existing bottlenecks and promote the industry to develop in a more intelligent and economical direction. Therefore, a low-cost autonomous inspection method and device of UAV for power overhead lines is urgently needed to solve the problems existing in the existing technology. Summary of the invention

[0006] The present invention aims to provide a low-cost autonomous inspection method for overhead power lines by drones, aiming to solve the problems of high cost and collision risk of existing autonomous inspection methods by drones. The specific technical solution is as follows: A low-cost autonomous inspection method for overhead power lines by drones, comprising: Obtain inspection images of the tower to be inspected; Use the target detection model to identify the type of tower in the inspection image and extract the 3D coordinate information of the tower; Establish a model library of various types of towers, and set the drone shooting points and flight paths for each type of tower; Based on the identified tower type, the tower's three-dimensional coordinate information and the model library, a parametric three-dimensional model of the identified tower is generated in the three-dimensional space map, and the drone shooting points and flight paths corresponding to the identified tower type are obtained; Analyze whether there is channel tree damage in each inspection image. If there is channel tree damage in the inspection image, optimize the drone shooting points and flight path of the tower in the inspection image; Generate the final drone autonomous inspection track file based on the optimized drone shooting points and flight paths and the drone shooting points and flight paths that do not need to be optimized .

[0007] Preferably, the specific method of analyzing whether there is a channel tree problem in the inspection image is: According to formula (2), Inspection image Convert from RGB color channels to HSV brightness space: (2), Among them, in formula (2) HSV Represents the conversion of RGB color channels into HSV Preprocessing operations in luminance space, represents the hue component, represents the saturation component, represents the brightness component, , Indicates the total number of inspection images; According to formula (3) and formula (4), the inspection image is statistically analyzed. The number of pixels belonging to green plants and trees ;like If it is greater than or equal to the set threshold, it is considered that there are green trees in the inspection image, that is, there is a tree hazard in the channel; If the value is less than the set threshold, it is considered that there are no green plants or trees in the inspection image, that is, there is no tree hazard in the passage; (3), (4), in, For inspection images The number of pixels belonging to green plants and trees counted in Indicates the inspection image middle , and The pixel value satisfies the formula (4).

[0008] Preferably, another method of analyzing whether there is a channel tree problem in the inspection image is: Extract inspection images Spectral characteristics of trees in G , for tree spectral characteristics G Perform multi-scale tree feature preservation filtering and combine the multi-scale filtering results to obtain ; Using principal component analysis Perform decision fusion and retain the first principal component ;right Perform threshold processing to generate a binary image , statistical binary image The number of pixels belonging to green plants and trees N ,like N If it is greater than or equal to the judgment threshold, the inspection image is considered There is a channel tree problem in , Indicates the total number of inspection images.

[0009] Preferably, the tree spectral characteristics are calculated according to formula (9): G Perform tree feature preserving filtering: (9), Representation domain transform recursive filtering; All in the Filter parameters under group parameter conditions; Indicated in Spectral characteristics of trees under different parameter conditions G The filtering result of According to formula (10), the multi-scale filtering results are combined to obtain : (10), in, Indicates the total number of groups of parameter conditions.

[0010] Preferably, generate the final drone autonomous inspection track file , specifically: (5), in, , and The value range of , Indicates the total number of inspection images. Indicates the number of inspection images with green plants and trees. Indicates the number of inspection images without green trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images with green plants and trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images without green plants and trees. Indicates the type of tower in the inspection image. , Indicates the total number of tower types.

[0011] Preferably, a drone is used to shoot above the pole tower to obtain an inspection image of the pole tower; wherein, during shooting, the pole tower target is ensured to be in the exact center of the inspection image.

[0012] Preferably, the target detection model is improved based on the yolov8 model, specifically: a Transformer layer is added after the SPPF layer in the yolov8 model backbone network, a BLRA layer is added in front of the SPPF layer in the yolov8 model backbone network, and the Conv convolution layer in the backbone network and the feature fusion network is replaced with a DSConv convolution layer.

[0013] Preferably, the power tower training data set Input into the target detection model to complete model training, use the trained target detection model to identify the type of tower in the inspection image, and extract the three-dimensional coordinate information of the tower; where: Respectively represent a tower type, Indicates the total number of tower types.

[0014] Preferably, the Inspection image Input the trained target detection model, then The recognition result of the inspection image is expressed as: (1), in, For the The recognition result of the inspection image is Indicates Inspection image Type of tower, , Indicates the total number of tower types, Indicates Inspection image The longitude and latitude of the middle tower, Indicates Inspection image The altitude information of the middle tower.

[0015] Preferably, the drone shooting points and flight paths corresponding to the pole towers with channel tree problems are optimized, specifically: based on the set drone shooting points and flight paths, the drone shooting points and flight paths of the pole towers are increased by H meters, where H is greater than 0.

[0016] The present invention also provides a low-cost electric power overhead line drone autonomous inspection device, comprising a memory and a processor. When the processor runs the computer instructions stored in the memory, it executes the low-cost electric power overhead line drone autonomous inspection method.

[0017] The present invention also provides another low-cost UAV autonomous inspection device for power overhead lines, comprising a path planning unit and a UAV, wherein the UAV is used to collect inspection images and inspect the overhead lines; the path planning unit executes the low-cost UAV autonomous inspection method for power overhead lines, and provides a final UAV autonomous inspection track file for the UAV inspection. .

[0018] The application of the technical solution of the present invention has the following beneficial effects: The present invention presets models of different types of power towers and drone shooting points and flight paths (i.e., model libraries), matches the tower models and drone shooting points and flight paths in the model library according to the identified tower types, and adaptively generates the final drone autonomous inspection track file in combination with the three-dimensional coordinate information of the identified towers; the present invention can achieve one-time investment and construction, and subsequent repeated applications, greatly reducing the application cost of drone autonomous inspections, and realizing low-cost distribution network drone autonomous inspections. In addition, the present invention qualitatively analyzes whether there are channel tree problems in the inspection image based on the number of green vegetation pixels in the inspection image, and then optimizes the drone inspection track file, avoids the risk of drones hitting trees, improves the safety of drone inspection operations, and improves the quality and efficiency of distribution network operation and maintenance. The present invention can realize low-cost drone autonomous inspections of power overhead lines under complex mountainous conditions, and has great significance and practical value in intelligent power operation and maintenance.

[0019] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a flow chart of the low-cost autonomous inspection method of power overhead line UAV in the present invention; Figure 2 It is a structural diagram of the target detection model in the present invention. DETAILED DESCRIPTION

[0021] In order to facilitate the understanding of the present invention, the present invention will be described more fully below, and preferred embodiments of the present invention are provided. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0023] Embodiment 1: See also Figure 1 This embodiment provides a low-cost autonomous inspection method for overhead power lines by drones, which is particularly suitable for complex mountainous areas where there are many dense trees and a high risk of collision during drone inspection. The inspection method includes the following steps: Step S1, obtaining an inspection image of a tower to be inspected; Specifically, in step S1, a drone is used to shoot above the pole tower to obtain an inspection image of the pole tower (ie, a visible light image); further, when shooting, it should be ensured that the pole tower target is in the exact center of the inspection image.

[0024] Step S2: using the target detection model to identify the type of tower in the inspection image, and extracting the three-dimensional coordinate information of the tower (i.e., longitude and latitude, altitude information); Specifically, the target detection model used in this embodiment is obtained based on the improvement of the yolov8 model. The improvement of the existing yolov8 model in this embodiment is as follows: a Transformer layer (i.e., encoder-decoder network module) is added after the SPPF layer (i.e., spatial pyramid pooling layer) in the backbone network of the yolov8 model, and a BLRA layer (i.e., Bi-level Routing Attention, double-layer routing attention mechanism module) is added in front of the SPPF layer in the backbone network of the yolov8 model. At the same time, the Conv convolution layer in the backbone network and the feature fusion network is replaced with a DSConv convolution layer (i.e., snake convolution layer). The improved yolov8 model improves the model calculation efficiency and performance through the BLRA layer, and the self-attention calculation and the use of the window attention mechanism in the local area of ​​the image are performed through the Transformer layer, which improves the efficiency of the model in processing image data and maintains a strong feature extraction capability. At the same time, the DSConv convolution layer is applied to realize the extraction of detailed information of the tower target at different spatial scales, thereby improving the accuracy of AI recognition.

[0025] See also Figure 2 , the target detection model in this embodiment is described in detail below: The target detection model in this embodiment includes a backbone network, a feature fusion network and a detection network, wherein the backbone network is used to extract multi-scale deep visual features from inspection images, the feature fusion network is used to fuse features of different scales, and the detection network is responsible for outputting detection information.

[0026] The backbone network includes DSConv convolution layer, DSConv convolution layer, 3×C2F layer (where C2f is a feature fusion layer), DSConv convolution layer, 6×C2F layer, DSConv convolution layer, 6×C2F layer, DSConv convolution layer, 3×C2F layer, BLRA layer, SPPF layer and Transformer layer in sequence; wherein: the DSConv convolution layer is used to dynamically adjust the shape and weight of the convolution kernel, which can better adapt to the diversity and complexity in the image, so as to extract richer and more accurate features; the 3×C2F layer and the 6×C2F layer are both used to splice the features of different branches in the channel dimension to achieve feature fusion; the SPPF layer is used to enhance the model for targets of different scales detection capability; the Transformer layer is used for self-attention calculation and the use of windowed attention mechanism to improve the efficiency of the model in processing image data; the BLRA layer is used to improve the computational efficiency and performance of the model. Specifically, the BLRA layer contains two attention mechanism channels. The upper branch attention mechanism channel interacts with all image blocks through the global self-attention mechanism and generates a global image representation; the lower branch attention mechanism channel uses the local self-attention mechanism to interact with each image block and its neighboring image blocks and generates a local image representation; through this two-layer routing attention mechanism, the improved yolov8 model can capture global and local feature information at the same time, thereby improving the performance of the model in visual tasks.

[0027] The feature fusion network includes UP layer (i.e., upsampling layer, Upsample layer), Concat layer (i.e., splicing layer), 3×C2F layer, UP layer, Concat layer, 3×C2F layer, DSConv convolution layer, Concat layer, 3×C2F layer, DSConv convolution layer, Concat layer and 3×C2F layer in sequence; wherein: the UP layer is an upsampling operation, which is used to fuse the information extracted from the deep feature map to the shallower feature map so that it can match the feature map with higher resolution; the Concat layer is used to merge feature maps from different depths to form a more comprehensive representation; the 3×C2F layer is used to splice the features of different branches in the channel dimension to achieve feature fusion.

[0028] The detection network includes three Detect layers (i.e., detection layers), and the last three 3×C2F layers of the feature fusion network are respectively connected to a Detect layer, and the Detect layer is used to determine the objects existing in the image and their positions.

[0029] Among them, the first 6×C2F layer in the backbone network is connected to the second Concat layer in the feature fusion network, the DSConv convolution layer between the 6×C2F layer and the 3×C2F layer in the backbone network (where the 6×C2F layer is located in front of the DSConv convolution layer and the 3×C2F layer is located behind the DSConv convolution layer) is connected to the first Concat layer in the feature fusion network, the Transformer layer in the backbone network is connected to the first UP layer and the last Concat layer in the feature fusion network, and the first 3×C2F layer in the feature fusion network is connected to the second to last Concat layer in the feature fusion network.

[0030] The processing process of the inspection image in the target detection model is as follows: The inspection image is input into the backbone network, and firstly extracted by two DSConv convolutional layers in sequence to obtain richer and more accurate feature data A. Then, feature data A is input into the 3×C2F layer (i.e., three C2F layers) for processing. Feature B output by the 3×C2F layer is processed by the DSConv convolutional layer and the 6×C2F layer (i.e., 6 C2F layers) in sequence, and feature Y1 and feature Y2 are output. Feature Y1 is processed by the DSConv convolutional layer, the 6×C2F layer, and the DSConv convolutional layer in sequence, and feature B1 and feature B2 are output. Feature B1 is processed by the 3×C2F layer, the BLRA layer, the SPPF layer, and the Transformer layer in sequence, and feature D1 and feature D2 are output. Further, feature Y2, feature B2, feature D1 and feature D2 are all input into the feature fusion network, feature D1 is input into the UP layer for processing and outputs feature E, feature E and feature B2 are concatenated through the Concat layer to output feature F, feature F is processed through the 3×C2F layer to output feature G1 and feature G2, feature G1 is processed through the UP layer and sent to the Concat layer for concatenation with feature Y2, and then input into the 3×C2F layer for processing and outputting feature K1 and feature K2, feature K1 is processed through the DSConv convolution layer and enters the Concat layer for concatenation with feature G2, and then input into the 3×C2F layer for processing and outputting feature L1 and feature L2, feature L1 is processed through the DSConv convolution layer and enters the Concat layer for concatenation with feature D2, and then input into the 3×C2F layer for processing and outputting feature M; Among them, feature K2 is sent to the first Detect layer to determine the objects and their positions in the image, feature L2 is sent to the second Detect layer to determine the objects and their positions in the image, and feature M is sent to the third Detect layer to determine the objects and their positions in the image.

[0031] In the processing of the model, the C2F layer performs feature conversion, branch processing and feature fusion operations on the input data through two convolutional layers (cv1 and cv2), extracts and converts the features of the input data, and generates a more representative feature output.

[0032] In order to adapt to the spatial morphological characteristics of the pole tower in this embodiment, the target detection model introduces a serpentine convolution layer (i.e., a serpentine convolution module) on the one hand. This module has better feature representation when processing slender tubular structures (such as slender targets such as pole towers); secondly, in the complex power inspection background, the ordinary YOLO algorithm is difficult to handle the problem of small target recognition (such as after the drone increases the altitude to shoot, the pole tower target will appear in the form of a small target). Therefore, by introducing the transformer module, performing self-attention calculation in the local area and using the window attention mechanism, the efficiency of the model in processing image data is improved, and the powerful feature extraction ability of the small target pole tower is maintained; finally, since the spatial resolution of the drone inspection image is very high, for example, the size of the picture taken by DJI is 5000×4000 pixels, and a picture has more than 20 megabytes, therefore, how to speed up the training and reasoning efficiency of the model is crucial to improving the inspection efficiency. In this embodiment, the BLRA attention mechanism module is introduced, and the dynamic and query-aware sparse attention mechanism is used to solve this problem, so that the target detection model of this embodiment can be well qualified for the task of power inspection image recognition.

[0033] Furthermore, the power tower training dataset Input into the target detection model to complete model training, where: Respectively represent a tower type, Indicates the total number of tower types. In this embodiment, five tower types are set. Represents a single-circuit straight tower. Indicates a portal type straight pole tower. Indicates the transformer tower. Indicates a double-circuit straight pole tower. represents a double-door tension tower, each tower type contains 2000 samples; during training, the core parameters are set as follows: batch size is set to 32, initial learning rate is set to 10 -3 , the optimizer is the Adam algorithm, label smoothing is set to 0.1, and the epoch is set to 120.

[0034] After the model training is completed, the tower type in the inspection image can be detected and identified. The inspection image is ,in , Represents the total number of inspection images input, then The recognition result of the inspection image is expressed as: (1), in, For the The recognition result of the inspection image is Indicates Inspection image Type of tower, , Indicates Inspection image The longitude and latitude of the middle tower, Indicates Inspection image The altitude information of the middle tower.

[0035] Step S3, establishing a model library of various types of towers, and setting the drone shooting points and flight paths for each type of tower; Specifically, those skilled in the art can set the drone shooting points and drone flight paths of various pole towers according to actual conditions. In this embodiment, the detailed setting of the drone shooting points and flight paths of various pole towers will not be described in detail. The shooting points and flight paths of the UAVs of the tower-like structures are recorded as , , Indicates the total number of tower types.

[0036] Step S4, based on the tower type identified in step S2, the three-dimensional coordinate information of the tower and the model library of step S3, a parameterized three-dimensional model of the identified tower is generated in the three-dimensional space map, and the drone shooting point and flight path corresponding to the identified tower type are obtained; Preferably, step S4 is specifically to obtain the three-dimensional model of the tower from the model library according to the type of the identified tower, determine the position of the tower in the three-dimensional space map based on the three-dimensional coordinate information of the identified tower, and after obtaining the parameterized three-dimensional model, pre-set the drone track file for each type of tower, that is, set the flight path and shooting points for shooting each type of tower, etc.; wherein the flight path plans the shooting order between the shooting points.

[0037] Step S5, analyzing whether there is a channel tree hazard in each inspection image, and if there is a channel tree hazard in the inspection image, optimizing the drone shooting point and flight path of the tower in the inspection image; Preferably, the specific method of analyzing whether there is a channel tree problem in the inspection image is: First, according to formula (2), the inspection image Convert from RGB color channels to HSV Luminance space to better handle colors in inspection images: (2), Among them, in formula (2) HSV Represents the conversion of RGB color channels into HSV Preprocessing operations in luminance space, represents the hue component, represents the saturation component, Represents the brightness component.

[0038] Then, according to formula (3) and formula (4), the inspection image is statistically analyzed. The number of pixels belonging to green plants and trees ;like If it is greater than or equal to the set threshold, it is considered that there are green trees in the inspection image, that is, there is a tree hazard in the channel; If the value is less than the set threshold, it is considered that there are no green plants or trees in the inspection image, that is, there is no tree hazard in the passage; (3), (4), in, For inspection images The number of pixels belonging to green plants and trees counted in Indicates the inspection image middle , and The pixel value satisfies the formula (4).

[0039] Preferably, the threshold value in this embodiment is set to 500 based on experience; of course, in other embodiments, the threshold value may be set to other values, and those skilled in the art can set it according to actual conditions. In this embodiment, the possible value range of the threshold value is no longer described in detail.

[0040] According to the standards for distribution network operation and maintenance, trees are not allowed under power poles and channels. Therefore, if green plants and trees are identified in the inspection image (such as trees growing under power poles and channels), in order to avoid unnecessary drone collisions, the drone shooting points and flight paths of the poles in the inspection image with green plants and trees need to be increased by H meters. In this embodiment, H is taken to be 5 meters. Those skilled in the art can determine the value of H based on actual terrain, experience, etc., and no detailed description is given in this embodiment.

[0041] Step S6: Generate the final UAV autonomous inspection track file ; In step S6, the drone shooting points and flight paths of some towers may be optimized, while the drone shooting points and flight paths of the remaining towers do not need to be optimized, so the final drone autonomous inspection track file It consists of two parts: (5), in, , and The value range of , Indicates the total number of inspection images input. Indicates the number of inspection images with green plants and trees. Indicates the number of inspection images without green trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images with green plants and trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images without green plants and trees. Indicates the type of tower in the inspection image. , Indicates the total number of tower types.

[0042] At this point, any drone only needs to import the defined final drone autonomous inspection track file , that is, autonomous inspection can be completed without any human intervention and processing.

[0043] This embodiment also provides a low-cost electric power overhead line autonomous inspection device by drone, including a memory and a processor, wherein the memory and the processor are connected, and when the processor runs the computer instructions stored in the memory, the low-cost electric power overhead line autonomous inspection method by drone is executed.

[0044] This embodiment also provides another low-cost UAV autonomous inspection device for power overhead lines, including a path planning unit and a UAV, wherein the UAV is used to collect inspection images and inspect the overhead lines; the path planning unit executes the low-cost UAV autonomous inspection method for power overhead lines, and provides a final UAV autonomous inspection track file for the UAV inspection. .

[0045] Embodiment 2: The difference between this embodiment and embodiment 1 is that the method of analyzing whether there is a channel tree problem in the inspection image is different. The specific method of analyzing whether there is a channel tree problem in the inspection image in this embodiment is: Extract inspection images Spectral characteristics of trees in G , for tree spectral characteristics G Perform multi-scale tree feature preservation filtering and combine the multi-scale filtering results to obtain ; Using principal component analysis Perform decision fusion and retain the first principal component ;right Perform threshold processing to generate a binary image , statistical binary image The number of pixels belonging to green plants and trees N ,like N If it is greater than or equal to the judgment threshold, the inspection image is considered There is a channel tree problem in the system.

[0046] Further, extract the inspection image Spectral characteristics of trees G The specific method is: According to formula (6), Inspection image Convert from RGB color channels to HSV brightness space: (6), Among them, in formula (6) HSV Represents the conversion of RGB color channels into HSV Preprocessing operations in luminance space, represents the hue component, represents the saturation component, represents the brightness component, , Indicates the total number of inspection images; According to formula (7) and formula (8), the inspection image is obtained Spectral characteristics of trees in G : (7), (8), in, Indicates the inspection image middle , and The pixel value satisfies the formula (8).

[0047] Although, through The function can detect the target pixels of most green trees, but the noise and isolated background in the image still interfere with the target. Therefore, in order to further improve the spectral characteristics of trees, G The present embodiment also analyzes the spectral characteristics of trees.G Perform multi-scale tree feature preservation filtering, and then combine the multi-scale filtering results to obtain , specifically: According to formula (9), the spectral characteristics of trees G Perform tree feature preserving filtering: (9), Representation domain transform recursive filtering; All in the Filter parameters under group parameter conditions, Set according to experience value; Indicated in Spectral characteristics of trees under different parameter conditions G The filtering result of According to formula (10), the multi-scale filtering results are combined to obtain : (10), in, Represents the total number of groups of parameter conditions (i.e., the number of scale types).

[0048] Furthermore, in order to fully explore the tree feature information under different filtering scales, this embodiment uses principal component analysis to Perform decision fusion processing and retain the first principal component : (11), in, Represents the principal component analysis dimensionality reduction operation.

[0049] Get fusion results Then perform threshold processing to generate a binary image ,exist The pixels with a pixel value of 1 indicate that they are most likely to be pixels of green plants and trees. In order to reduce the misjudgment rate, a judgment threshold is set in this embodiment. N When it is greater than or equal to the judgment threshold, it is considered that there are tree hazards in the inspection image; among them, the statistical binary image The operation of the pixels belonging to green plants and trees is expressed as: (12), in, num This is an operation to count the number of pixels.

[0050] In this embodiment, the determination threshold is set to 500; compared with the tree damage identification method in Example 1, the identification method in this embodiment can avoid the interference of noise and isolated background in the image, improve the accuracy of tree damage identification, and avoid misjudgment.

[0051] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A low-cost autonomous inspection method for overhead power lines by drones, characterized in that: include: Obtain inspection images of the tower to be inspected; Use the target detection model to identify the type of tower in the inspection image and extract the 3D coordinate information of the tower; Establish a model library of various types of towers, and set the drone shooting points and flight paths for each type of tower; Based on the identified tower type, the tower's three-dimensional coordinate information and the model library, a parametric three-dimensional model of the identified tower is generated in the three-dimensional space map, and the drone shooting points and flight paths corresponding to the identified tower type are obtained; Analyze whether there is channel tree damage in each inspection image. If there is channel tree damage in the inspection image, optimize the drone shooting points and flight path of the tower in the inspection image; Generate the final drone autonomous inspection track file based on the optimized drone shooting points and flight paths and the drone shooting points and flight paths that do not need to be optimized .

2. The low-cost autonomous inspection method for overhead power lines by drones according to claim 1 is characterized in that: The specific method to analyze whether there is channel tree problem in the inspection image is as follows: According to formula (2), Inspection image Convert from RGB color channels to HSV brightness space: （2）， Among them, in formula (2) HSV Represents the conversion of RGB color channels into HSV Preprocessing operations in luminance space, represents the hue component, represents the saturation component, represents the brightness component, , Indicates the total number of inspection images; According to formula (3) and formula (4), the inspection image is statistically analyzed. The number of pixels belonging to green plants and trees ;like If it is greater than or equal to the set threshold, it is considered that there are green trees in the inspection image, that is, there is a tree hazard in the channel; If the value is less than the set threshold, it is considered that there are no green plants or trees in the inspection image, that is, there is no tree hazard in the passage; （3）， （4）， in, For inspection images The number of pixels belonging to green plants and trees counted in Indicates the inspection image middle , and The pixel value satisfies the formula (4).

3. The low-cost autonomous inspection method for overhead power lines by drone according to claim 1 is characterized in that: The specific method to analyze whether there is channel tree problem in the inspection image is as follows: Extract inspection images Spectral characteristics of trees in G , for tree spectral characteristics G Perform multi-scale tree feature preservation filtering and combine the multi-scale filtering results to obtain ; Using principal component analysis Perform decision fusion and retain the first principal component ;right Perform threshold processing to generate a binary image , statistical binary image The number of pixels belonging to green plants and trees N ,like N If it is greater than or equal to the judgment threshold, the inspection image is considered There is a channel tree problem in , Indicates the total number of inspection images.

4. The low-cost autonomous inspection method for overhead power lines by drone according to claim 3 is characterized in that: According to formula (9), the spectral characteristics of trees G Perform tree feature preserving filtering: （9）， Representation domain transform recursive filtering; All in the Filter parameters under group parameter conditions; Indicated in Spectral characteristics of trees under different parameter conditions G The filtering result of According to formula (10), the multi-scale filtering results are combined to obtain : （10）， in, Indicates the total number of groups of parameter conditions.

5. The low-cost autonomous inspection method for overhead power lines by drones according to any one of claims 2 to 4, characterized in that: Generate the final drone autonomous inspection track file , specifically: （5）， in, , and The value range of , Indicates the total number of inspection images. Indicates the number of inspection images with green plants and trees. Indicates the number of inspection images without green trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images with green plants and trees. Indicates The drone shooting points and flight paths of the towers corresponding to the inspection images without green plants and trees. Indicates the type of tower in the inspection image. , Indicates the total number of tower types.

6. The low-cost autonomous inspection method for overhead power lines by drone according to claim 1 is characterized in that: The target detection model is improved based on the yolov8 model, specifically: a Transformer layer is added after the SPPF layer in the yolov8 model backbone network, a BLRA layer is added before the SPPF layer in the yolov8 model backbone network, and the Conv convolution layer in the backbone network and the feature fusion network is replaced with a DSConv convolution layer.

7. The low-cost autonomous inspection method for overhead power lines by drone according to claim 1 is characterized in that: The first Inspection image Input the trained target detection model, then The recognition result of the inspection image is expressed as: （1）， in, For the The recognition result of the inspection image. Indicates Inspection image Type of tower, , Indicates the total number of tower types, Indicates Inspection image The longitude and latitude of the middle tower, Indicates Inspection image The altitude information of the middle tower.

8. The low-cost autonomous inspection method for overhead power lines by drone according to claim 1 is characterized in that: The drone shooting points and flight paths corresponding to the towers with channel tree problems are optimized. Specifically, based on the set drone shooting points and flight paths, the drone shooting points and flight paths of the towers are increased by H meters, where H is greater than 0.

9. A low-cost UAV autonomous inspection device for power overhead lines, characterized in that: It comprises a memory and a processor, and when the processor runs the computer instructions stored in the memory, it executes the low-cost autonomous inspection method of overhead power lines by unmanned aerial vehicles as described in any one of claims 1 to 8.

10. A low-cost UAV autonomous inspection device for power overhead lines, characterized in that: The invention comprises a path planning unit and a drone, wherein the drone is used to collect inspection images and inspect the overhead lines; the path planning unit executes the low-cost autonomous inspection method for overhead power lines by drones as described in any one of claims 1 to 8, and provides a final autonomous inspection track file for the drone inspection. .

Images