Unmanned aerial vehicle system for transformer substation inspection

Abstract

The invention discloses an unmanned aerial vehicle system for transformer substation inspection, and relates to the technical field of transformer substation inspection, an inspection task unit is arranged to obtain an inspection task, a panoramic image unit is used to obtain a panoramic image of a transformer substation, and a raster map unit is arranged to construct a raster map. Routing inspection path planning is carried out on the grid map through the routing inspection path unit; the unmanned aerial vehicle unit executes an inspection task based on the inspection path; the image processing unit processes the collected image data in real time, if it is judged that a fault exists in a corresponding area according to a matching result, fault marking is carried out on the corresponding area through the adjusting unit, and if it is judged that an inspection missing area exists in the inspected area according to the matching result, the unmanned aerial vehicle is adjusted to reinspect the inspection missing area. According to the invention, the inspection path is optimized, the overall inspection path of the unmanned aerial vehicle is improved, and the collected data is processed in real time, so that a fault area can be marked and an undetected area can be rechecked in time.

Description

Technical Field

[0001] The present invention relates to the technical field of substation inspection, and in particular to an unmanned aerial vehicle system for substation inspection. Background Art

[0002] The substation inspection drone system is a solution that uses drone technology to conduct automated and intelligent inspections of substations. In the existing technology, the principle of using drones to inspect substations is: the operation and maintenance personnel pre-plan the flight path and inspection tasks of the drone through the background software, set the flight altitude, speed and shooting parameters; the drone flies along the preset path and uses a variety of sensors to collect images and data of substation equipment; the collected data is transmitted to the ground control station or cloud server through the wireless communication module; the background software processes and analyzes the collected data to generate a high-precision three-dimensional model and inspection report; the operation and maintenance personnel perform on-site processing based on the inspection report to ensure the normal operation of the equipment.

[0003] However, the above scheme has the following disadvantages: 1) In path planning, the common method is to set up multiple straight flight paths, and the adjacent straight flight paths are connected end to end. Although this path planning can cover all areas of the substation, its inspection path is not the optimal path, resulting in low overall inspection efficiency; 2) In the existing technology, the drone transmits the collected data to the ground control station or cloud server in real time, but the collected data needs to be uniformly processed and analyzed by the background after the inspection task is completed to generate an inspection report. Considering that the amount of collected data is often large, when a problem occurs in the substation, it is often not discovered and handled in the first time through drone inspection; 3) During the drone inspection process, missed inspections may occur. The current processing method is to process the collected data after the inspection task is completed, find the missed inspection area, and use the drone to fly again or manually inspect the area to solve the problem, which further reduces the drone inspection efficiency.

[0004] In order to solve the above problems, there is an urgent need for a drone system for substation inspection. Summary of the invention

[0005] In order to solve the above-mentioned deficiencies in the prior art, the present invention provides a drone system for substation inspection, the system comprising: Inspection task unit: obtain inspection tasks; Panoramic image unit: using multiple cameras to capture local images of the substation to be inspected from multiple angles, and processing all local images to obtain a panoramic image of the substation to be inspected; Grid map unit: Based on the substation cloud data scanned by 3D LiDAR, a grid map is constructed for planning the inspection path of drones. Inspection path unit: based on the inspection task, marking the inspection starting point, multiple inspection nodes and the inspection end point in the grid map, and based on the marked grid map, adopting the first method to perform UAV inspection path planning; UAV unit: based on the planned UAV inspection path, executes the inspection task and transmits the collected image data back to the image processing unit in real time; An image processing unit is used to process the collected image data to obtain a real-time image of the substation to be inspected, match the real-time image with the panoramic image once, and obtain a first matching result; use a second method to stitch all the real-time images to obtain a stitched image of the substation to be inspected, and match the stitched image with the panoramic image twice to obtain a second matching result; Adjustment unit: Based on the first matching result, determine whether a fault occurs in the corresponding area of ​​the real-time image, and if so, mark the corresponding area for fault; based on the second matching result, determine whether there is a missed inspection area in the area inspected by the drone, and if so, use a third method to adjust the drone.

[0006] The present invention is implemented through the following technical scheme: this scheme first obtains the corresponding inspection task through the inspection task unit. In addition to daily inspections, the substation also includes key inspections on certain equipment. Therefore, it is necessary to set corresponding inspection paths for different inspection tasks; a panoramic image unit is used to obtain a panoramic image of the substation; a grid map unit is constructed, and point cloud data of the substation is obtained through a three-dimensional laser radar, and the equipment distribution data of the substation is converted to a grid map; the inspection path unit is used to plan the inspection path of the drone on the grid map. First, the inspection starting point, inspection node and inspection end point to be traversed by the drone are determined through the inspection task, and then the shortest inspection path of the drone is planned on the grid map through the first method; the drone unit is used to execute the inspection task based on the inspection path, and during the inspection process The collected image data is transmitted back to the image processing unit in real time; the image processing unit is used to process the collected image data in real time, firstly, the real-time image is extracted from the transmitted video stream data, the real-time image is matched with the panoramic image once to obtain a first matching result, and then all the extracted real-time images are spliced ​​to obtain a spliced ​​image by a second method, and then the spliced ​​image is matched with the panoramic image for a second time to obtain a second matching result; the adjustment unit is used to determine whether a fault occurs in the corresponding area of ​​the real-time image based on the first matching result, and if so, the corresponding area is marked for fault, and based on the second matching result, it is determined whether there is a missed inspection area in the area inspected by the drone, and if so, the third method is used to adjust the drone so that the drone can re-inspect the missed inspection area in time during the inspection process.

[0007] As an optional technical solution, all local images are processed to obtain a panoramic image of the substation to be inspected, including: Get all partial images and put them into the preset image set; Acquire any two partial images with overlapping areas from the preset image set and process them using a preset method, wherein the any two partial images with overlapping areas are recorded as a first image and a second image, and the preset method includes: Extracting a plurality of first key feature points in the first image and a plurality of second key feature points in the second image using a SIFT algorithm; Matching all first key feature points in the first image with all second key feature points in the second image in sequence to obtain a third matching result; Based on the third matching result, a first homography matrix is ​​calculated; Based on the first homography matrix, transform the second image into the coordinate system of the first image, and perform perspective transformation on the second image and then splice it with the first image; removing a black area after the first image and the second image are spliced ​​together, and outputting a spliced ​​image of the first image and the second image; The preset method is executed on all the local images with overlapping areas in the preset image set until a panoramic image of the substation to be inspected is generated by splicing.

[0008] As an optional technical solution, based on the substation cloud data scanned by 3D LiDAR, a grid map for planning drone inspection paths is constructed, including: Obtain all substation cloud data based on 3D LiDAR scanning and put it into the preset data set; Determine whether the height value of any substation cloud data in the preset data set is less than a first threshold value, if so, determine that the corresponding substation cloud data is a ground point and filter it out once, if not, do nothing; Setting a valid value range for the substation site cloud data, and performing a second filtering on the preset data set after the first filtering based on the valid value range; Creating a three-dimensional grid space based on all substation cloud data in the preset data set after secondary filtering; An initial grid map is constructed, and the points corresponding to the cloud data of all substations in the three-dimensional grid space are projected onto the initial grid map. An expansion coefficient is set for the points corresponding to the substation equipment on the initial grid map to form a grid map for planning the inspection path of drones.

[0009] As an optional technical solution, the first method includes: Setting a first set to store expanded nodes, and a second set to store nodes to be expanded; Put the inspection starting point into the first set, and put all inspection nodes and the inspection end point into the second set; The first formula is used to calculate the cost value of the inspection starting point and any inspection node, including: is the cost from the inspection starting point to any inspection node, Indicates the actual cost from the inspection starting point to any inspection node. is the weight of the actual cost, It represents the estimated cost from the inspection starting point to any inspection node. represents the weight of the estimated cost, Indicates the search tendency cost from the inspection starting point to any inspection node. The weight representing the cost of search tendency; Based on the first formula, the adjacent inspection nodes of the inspection starting point are screened in the second set, and the adjacent inspection nodes are used as new inspection starting points, and the adjacent inspection nodes of the new inspection starting point are screened in the second set until the adjacent inspection nodes are the inspection end points.

[0010] As an optional technical solution, the real-time image is matched with the panoramic image once by using the second method, and the first matching result is obtained, including: Extracting a plurality of first feature points of the real-time image and a plurality of second feature points in the panoramic image, and calculating descriptors of all first feature points, recorded as first descriptors, and descriptors of all second feature points, recorded as second descriptors; Calculating the Euclidean distances between all first descriptors and all second descriptors, and searching for the nearest neighbor feature points of all first feature points of the real-time image in the panoramic image; A second homography matrix is ​​estimated from all nearest neighbor feature points, and based on the second homography matrix, the real-time image is fused into the panoramic image to obtain the first matching result.

[0011] As an optional technical solution, all real-time images are spliced ​​to obtain a spliced ​​image of the substation to be inspected, including: All real-time images are binarized and converted into grayscale images; Each grayscale image is divided into an overlapping area and a non-overlapping area, and the grayscale difference value between the overlapping areas in any two grayscale images is calculated to obtain a calculation result; Based on the calculation results, any two grayscale images are adjusted so that the edges of corresponding overlapping areas match each other, and any two adjusted grayscale images in all real-time images are spliced ​​to obtain a spliced ​​image of the substation to be inspected.

[0012] As an optional technical solution, performing secondary matching on the stitched image and the panoramic image to obtain a second matching result includes: Extracting a plurality of third feature points from the stitched image and a plurality of fourth feature points from the panoramic image, and calculating descriptors of all third feature points, recorded as third descriptors, and descriptors of all fourth feature points, recorded as fourth descriptors; Calculating the Euclidean distances between all third descriptors and all fourth descriptors, and searching for the nearest neighbor feature points of all third feature points of the real-time image in the panoramic image; A third homography matrix is ​​estimated from all nearest neighbor feature points, and based on the third homography matrix, the stitched image is fused into the panoramic image to obtain the second matching result.

[0013] As an optional technical solution, the third method includes: When it is determined that there is a missed inspection area in the area that the drone has inspected, it is recorded as the first missed inspection area; Obtain the coordinates of the current location of the drone, recorded as the first coordinate point, and the coordinates of the center position of the first missed detection area, recorded as the second coordinate point; Based on the first coordinate point and the second coordinate point, planning a secondary inspection path of the drone from the current position to the first missed inspection area; Get the current flight altitude of the drone, recorded as the first altitude; Based on the first height, determine whether there are obstacles on the secondary inspection path. If so, use the fourth method to adjust the secondary inspection path. If not, the drone re-inspects the first missed inspection area based on the secondary inspection path.

[0014] As an optional technical solution, the fourth method includes: When it is determined that there is an obstacle on the secondary inspection path, it is recorded as the first obstacle; Acquire and model the first obstacle based on the three-dimensional point cloud data of the first obstacle; The A* algorithm is used to calculate the shortest path for the drone to pass through the first obstacle.

[0015] One or more technical solutions provided by the present invention have at least the following technical effects or advantages: The invention discloses a UAV system for substation inspection, wherein the inspection task is obtained by setting an inspection task unit, the panoramic image of the substation is obtained by a panoramic image unit, the distribution data of the substation is converted to a grid map by setting a grid map unit, and the inspection path unit is used to plan the inspection path on the grid map to obtain the shortest inspection path of the UAV; the UAV unit then performs the corresponding inspection task based on the planned inspection path, and collects image data during the inspection process; the image processing unit is used to process the collected image data in real time, firstly, the real-time image is matched with the panoramic image once, if it is judged that the corresponding area has a fault according to the matching result, the corresponding area is marked by the adjustment unit, and then the real-time image is spliced ​​to obtain a spliced ​​image, and the spliced ​​image is matched with the panoramic image for a second time, if it is judged that there is a missed inspection area in the inspected area according to the matching result, the UAV is adjusted to re-inspect the missed inspection area. The invention not only improves the overall inspection path of the UAV by optimizing the inspection path, but also processes the collected data in real time, and can timely mark the fault area and re-inspect the missed inspection area. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention; Figure 1 It is a schematic diagram of the composition of a drone system used for substation inspection in the present invention. DETAILED DESCRIPTION

[0017] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0019] Example Please refer to Figure 1 , Figure 1 The invention relates to a drone system for substation inspection, the system comprising: Inspection task unit: obtain inspection tasks; Panoramic image unit: using multiple cameras to capture local images of the substation to be inspected from multiple angles, and processing all local images to obtain a panoramic image of the substation to be inspected; Grid map unit: Based on the substation cloud data scanned by 3D LiDAR, a grid map is constructed for planning the inspection path of drones. Inspection path unit: based on the inspection task, marking the inspection starting point, multiple inspection nodes and the inspection end point in the grid map, and based on the marked grid map, adopting the first method to perform UAV inspection path planning; UAV unit: based on the planned UAV inspection path, executes the inspection task and transmits the collected image data back to the image processing unit in real time; An image processing unit is used to process the collected image data to obtain a real-time image of the substation to be inspected, match the real-time image with the panoramic image once, and obtain a first matching result; use a second method to stitch all the real-time images to obtain a stitched image of the substation to be inspected, and match the stitched image with the panoramic image twice to obtain a second matching result; Adjustment unit: Based on the first matching result, determine whether a fault occurs in the corresponding area of ​​the real-time image, and if so, mark the corresponding area for fault; based on the second matching result, determine whether there is a missed inspection area in the area inspected by the drone, and if so, use a third method to adjust the drone.

[0020] The specific embodiments of the present invention are as follows: Inspection task unit: Used to obtain inspection tasks. In addition to daily inspections, substations also include regular inspections and key inspections of special equipment. Therefore, in order to ensure the inspection efficiency of drones, corresponding inspection routes are planned for each inspection task; Panoramic image unit: Processing all local images to obtain a panoramic image of the substation to be inspected includes: Get all partial images and put them into the preset image set; Acquire any two partial images with overlapping areas from the preset image set and process them using a preset method, wherein the any two partial images with overlapping areas are recorded as a first image and a second image, and the preset method includes: Extracting a plurality of first key feature points in the first image and a plurality of second key feature points in the second image using a SIFT algorithm; Matching all first key feature points in the first image with all second key feature points in the second image in sequence to obtain a third matching result; Based on the third matching result, a first homography matrix is ​​calculated; Based on the first homography matrix, transform the second image into the coordinate system of the first image, and perform perspective transformation on the second image and then splice it with the first image; removing a black area after the first image and the second image are spliced ​​together, and outputting a spliced ​​image of the first image and the second image; The preset method is executed on all the local images with overlapping areas in the preset image set until a panoramic image of the substation to be inspected is generated by splicing.

[0021] Raster map unit: Based on the substation cloud data scanned by 3D LiDAR, a raster map for planning drone inspection paths is constructed, including: Obtain all substation cloud data based on 3D LiDAR scanning and put it into the preset data set; Determine whether the height value of any substation cloud data in the preset data set is less than a first threshold value, if so, determine that the corresponding substation cloud data is a ground point and filter it out once, if not, do nothing; Setting a valid value range for the substation site cloud data, and performing a second filtering on the preset data set after the first filtering based on the valid value range; Creating a three-dimensional grid space based on all substation cloud data in the preset data set after secondary filtering; An initial grid map is constructed, and the points corresponding to the cloud data of all substations in the three-dimensional grid space are projected onto the initial grid map. An expansion coefficient is set for the points corresponding to the substation equipment on the initial grid map to form a grid map for planning the inspection path of drones.

[0022] Among them, a grid map is constructed to reflect the equipment distribution of the substation. In the process of constructing the raster map, other data that affect the accuracy of the drone path planning are filtered out by filtering the ground point cloud and the effective value of the point cloud. The expansion coefficient of the substation equipment is introduced into the raster map. First, the safety of the drone during the flight can be guaranteed as much as possible, and second, the inspection path planning can be more reasonable.

[0023] Inspection path unit: Based on the inspection task, an inspection starting point, multiple inspection nodes and an inspection end point are marked in the grid map, and based on the marked grid map, a first method is used to plan the inspection path of the drone; The first method comprises: Setting a first set to store expanded nodes, and a second set to store nodes to be expanded; Put the inspection starting point into the first set, and put all inspection nodes and the inspection end point into the second set; The first formula is used to calculate the cost value of the inspection starting point and any inspection node, including: is the cost from the inspection starting point to any inspection node, Indicates the actual cost from the inspection starting point to any inspection node. is the weight of the actual cost, It represents the estimated cost from the inspection starting point to any inspection node. represents the weight of the estimated cost, Indicates the search tendency cost from the inspection starting point to any inspection node. The weight representing the search tendency cost; Based on the first formula, the adjacent inspection nodes of the inspection starting point are screened in the second set, and the adjacent inspection nodes are used as new inspection starting points, and the adjacent inspection nodes of the new inspection starting point are screened in the second set until the adjacent inspection nodes are the inspection end points.

[0024] Among them, this embodiment sets the first set to store the expanded nodes, sets the second set to store the nodes to be expanded, and puts the inspection starting point into the first set, and then uses the first formula to calculate the inspection node with the lowest cost with the inspection starting point in the second set as the adjacent node of the inspection starting point, and uses the adjacent node as the new inspection starting point, and searches for the next adjacent node in the second set until the adjacent node is the inspection end point. It should be noted that the inspection nodes that are successfully screened in the second set will be placed in the first set and will not be retained in the second set.

[0025] Drone Unit: Based on the planned UAV inspection path, the inspection task is performed, and the collected image data is transmitted back to the image processing unit in real time; the size and type of the UAV can be adjusted according to actual conditions, and this embodiment does not make specific limitations.

[0026] Image processing unit: Used to process the collected image data to obtain a real-time image of the substation to be inspected, match the real-time image with the panoramic image once, and obtain a first matching result; use a second method to stitch all the real-time images to obtain a stitched image of the substation to be inspected, and match the stitched image with the panoramic image twice to obtain a second matching result; The real-time image and the panoramic image are matched once using the second method to obtain a first matching result, including: Extracting a plurality of first feature points of the real-time image and a plurality of second feature points in the panoramic image, and calculating descriptors of all first feature points, recorded as first descriptors, and descriptors of all second feature points, recorded as second descriptors; Calculating the Euclidean distances between all first descriptors and all second descriptors, and searching for the nearest neighbor feature points of all first feature points of the real-time image in the panoramic image; A second homography matrix is ​​estimated from all nearest neighbor feature points, and based on the second homography matrix, the real-time image is fused into the panoramic image to obtain the first matching result.

[0027] All real-time images are stitched together to obtain a stitched image of the substation to be inspected, including: All real-time images are binarized and converted into grayscale images; Each grayscale image is divided into an overlapping area and a non-overlapping area, and the grayscale difference value between the overlapping areas in any two grayscale images is calculated to obtain a calculation result; Based on the calculation results, any two grayscale images are adjusted so that the edges of corresponding overlapping areas match each other, and any two adjusted grayscale images in all real-time images are spliced ​​to obtain a spliced ​​image of the substation to be inspected.

[0028] Performing secondary matching between the stitched image and the panoramic image to obtain a second matching result includes: Extracting a plurality of third feature points from the stitched image and a plurality of fourth feature points from the panoramic image, and calculating descriptors of all third feature points, recorded as third descriptors, and descriptors of all fourth feature points, recorded as fourth descriptors; Calculating the Euclidean distances between all third descriptors and all fourth descriptors, and searching for the nearest neighbor feature points of all third feature points of the real-time image in the panoramic image; A third homography matrix is ​​estimated from all nearest neighbor feature points, and based on the third homography matrix, the stitched image is fused into the panoramic image to obtain the second matching result.

[0029] Among them, the image processing unit is used to perform real-time processing on the collected image data. First, a real-time image is extracted from the returned video stream data, and the real-time image is matched with the panoramic image once to obtain a first matching result. Then, a second method is used to splice all the extracted real-time images to obtain a spliced ​​image, and then the spliced ​​image is matched with the panoramic image for a second time to obtain a second matching result.

[0030] Adjustment unit: Based on the first matching result, determine whether a fault occurs in the corresponding area of ​​the real-time image. If so, mark the corresponding area for fault. Based on the second matching result, determine whether there is a missed inspection area in the area inspected by the drone. If so, use the third method to adjust the drone.

[0031] The third method comprises: When it is determined that there is a missed inspection area in the area that the drone has inspected, it is recorded as the first missed inspection area; Obtain the coordinates of the current location of the drone, recorded as the first coordinate point, and the coordinates of the center position of the first missed detection area, recorded as the second coordinate point; Based on the first coordinate point and the second coordinate point, planning a secondary inspection path of the drone from the current position to the first missed inspection area; Get the current flight altitude of the drone, recorded as the first altitude; Based on the first height, determine whether there are obstacles on the secondary inspection path. If so, use the fourth method to adjust the secondary inspection path. If not, the drone re-inspects the first missed inspection area based on the secondary inspection path.

[0032] The fourth method comprises: When it is determined that there is an obstacle on the secondary inspection path, it is recorded as the first obstacle; Acquire and model the first obstacle based on the three-dimensional point cloud data of the first obstacle; The A* algorithm is used to calculate the shortest path for the drone to pass through the first obstacle.

[0033] Among them, the adjustment unit first determines whether a fault occurs in the corresponding area of ​​the real-time image based on the first matching result. If so, the corresponding area is marked for fault, and then based on the second matching result, it is determined whether there is a missed inspection area in the area inspected by the drone. If so, the position coordinates of the current drone are obtained, recorded as the first coordinate point, and the position coordinates of the missed inspection area are recorded as the second coordinate point. Based on the first coordinate point and the second coordinate point, the drone is planned to have a secondary inspection path from the current position to the missed inspection area, and by obtaining the flight altitude of the drone, it is determined whether there is an obstacle (i.e., substation equipment) on the secondary inspection path. If so, the secondary inspection path needs to be adjusted, and the obstacle is modeled and the A* algorithm is used to calculate the shortest path for the drone to pass through the obstacle. If not, the drone re-inspects based on the re-inspection path. It should be noted that after the drone re-inspects the missed inspection area, it returns to the current position along the original route and performs subsequent inspection tasks.

[0034] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0035] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A drone system for substation inspection, characterized in that: The system comprises: Inspection task unit: obtain inspection tasks; Panoramic image unit: using multiple cameras to capture local images of the substation to be inspected from multiple angles, and processing all local images to obtain a panoramic image of the substation to be inspected; Grid map unit: Based on the substation cloud data scanned by 3D LiDAR, a grid map is constructed for planning the inspection path of drones. Inspection path unit: based on the inspection task, marking the inspection starting point, multiple inspection nodes and the inspection end point in the grid map, and based on the marked grid map, adopting the first method to perform UAV inspection path planning; UAV unit: based on the planned UAV inspection path, executes the inspection task and transmits the collected image data back to the image processing unit in real time; An image processing unit is used to process the collected image data to obtain a real-time image of the substation to be inspected, match the real-time image with the panoramic image once, and obtain a first matching result; use a second method to stitch all the real-time images to obtain a stitched image of the substation to be inspected, and match the stitched image with the panoramic image twice to obtain a second matching result; Adjustment unit: Based on the first matching result, determine whether a fault occurs in the corresponding area of ​​the real-time image, and if so, mark the corresponding area for fault; based on the second matching result, determine whether there is a missed inspection area in the area inspected by the drone, and if so, use a third method to adjust the drone.

2. The UAV system for substation inspection according to claim 1 is characterized in that: Processing all local images to obtain a panoramic image of the substation to be inspected includes: Get all partial images and put them into the preset image set; Acquire any two partial images with overlapping areas from the preset image set and process them using a preset method, wherein the any two partial images with overlapping areas are recorded as a first image and a second image, and the preset method includes: Extracting a plurality of first key feature points in the first image and a plurality of second key feature points in the second image using a SIFT algorithm; Matching all first key feature points in the first image with all second key feature points in the second image in sequence to obtain a third matching result; Based on the third matching result, a first homography matrix is ​​calculated; Based on the first homography matrix, transform the second image into the coordinate system of the first image, and perform perspective transformation on the second image and then splice it with the first image; removing a black area after the first image and the second image are spliced ​​together, and outputting a spliced ​​image of the first image and the second image; The preset method is executed on all the local images with overlapping areas in the preset image set until a panoramic image of the substation to be inspected is generated by splicing.

3. The UAV system for substation inspection according to claim 1 is characterized in that: Based on the substation cloud data scanned by 3D LiDAR, a raster map for planning drone inspection paths is constructed, including: Obtain all substation cloud data based on 3D LiDAR scanning and put it into the preset data set; Determine whether the height value of any substation cloud data in the preset data set is less than a first threshold value, if so, determine that the corresponding substation cloud data is a ground point and filter it out once, if not, do nothing; Setting a valid value range for the substation site cloud data, and performing a second filtering on the preset data set after the first filtering based on the valid value range; Creating a three-dimensional grid space based on all substation cloud data in the preset data set after secondary filtering; An initial grid map is constructed, and the points corresponding to the cloud data of all substations in the three-dimensional grid space are projected onto the initial grid map. An expansion coefficient is set for the points corresponding to the substation equipment on the initial grid map to form a grid map for planning the inspection path of drones.

4. The UAV system for substation inspection according to claim 1 is characterized in that: The first method comprises: Setting a first set to store expanded nodes, and a second set to store nodes to be expanded; Put the inspection starting point into the first set, and put all inspection nodes and the inspection end point into the second set; The first formula is used to calculate the cost value of the inspection starting point and any inspection node, including: is the cost from the inspection starting point to any inspection node, Indicates the actual cost from the inspection starting point to any inspection node. is the weight of the actual cost, It represents the estimated cost from the inspection starting point to any inspection node. represents the weight of the estimated cost, Indicates the search tendency cost from the inspection starting point to any inspection node. The weight representing the cost of search tendency; Based on the first formula, the adjacent inspection nodes of the inspection starting point are screened in the second set, and the adjacent inspection nodes are used as new inspection starting points, and the adjacent inspection nodes of the new inspection starting point are screened in the second set until the adjacent inspection nodes are the inspection end points.

5. The UAV system for substation inspection according to claim 1 is characterized in that: The real-time image and the panoramic image are matched once using the second method to obtain a first matching result, including: Extracting a plurality of first feature points of the real-time image and a plurality of second feature points in the panoramic image, and calculating descriptors of all first feature points, recorded as first descriptors, and descriptors of all second feature points, recorded as second descriptors; Calculating the Euclidean distances between all first descriptors and all second descriptors, and searching for the nearest neighbor feature points of all first feature points of the real-time image in the panoramic image; A second homography matrix is ​​estimated from all nearest neighbor feature points, and based on the second homography matrix, the real-time image is fused into the panoramic image to obtain the first matching result.

6. The UAV system for substation inspection according to claim 1, characterized in that: All real-time images are stitched together to obtain a stitched image of the substation to be inspected, including: All real-time images are binarized and converted into grayscale images; Each grayscale image is divided into an overlapping area and a non-overlapping area, and the grayscale difference value between the overlapping areas in any two grayscale images is calculated to obtain a calculation result; Based on the calculation results, any two grayscale images are adjusted so that the edges of corresponding overlapping areas match each other, and any two adjusted grayscale images in all real-time images are spliced ​​to obtain a spliced ​​image of the substation to be inspected.

7. The UAV system for substation inspection according to claim 1 is characterized in that: Performing secondary matching between the stitched image and the panoramic image to obtain a second matching result includes: Extracting a plurality of third feature points from the stitched image and a plurality of fourth feature points from the panoramic image, and calculating descriptors of all third feature points, recorded as third descriptors, and descriptors of all fourth feature points, recorded as fourth descriptors; Calculating the Euclidean distances between all third descriptors and all fourth descriptors, and searching for the nearest neighbor feature points of all third feature points of the real-time image in the panoramic image; A third homography matrix is ​​estimated from all nearest neighbor feature points, and based on the third homography matrix, the stitched image is fused into the panoramic image to obtain the second matching result.

8. The UAV system for substation inspection according to claim 1, characterized in that: The third method comprises: When it is determined that there is a missed inspection area in the area that the drone has inspected, it is recorded as the first missed inspection area; Obtain the coordinates of the current location of the drone, recorded as the first coordinate point, and the coordinates of the center position of the first missed detection area, recorded as the second coordinate point; Based on the first coordinate point and the second coordinate point, planning a secondary inspection path of the drone from the current position to the first missed inspection area; Get the current flight altitude of the drone, recorded as the first altitude; Based on the first height, determine whether there are obstacles on the secondary inspection path. If so, use the fourth method to adjust the secondary inspection path. If not, the drone re-inspects the first missed inspection area based on the secondary inspection path.

9. The UAV system for substation inspection according to claim 8, characterized in that: The fourth method comprises: When it is determined that there is an obstacle on the secondary inspection path, it is recorded as the first obstacle; Acquire and model the first obstacle based on the three-dimensional point cloud data of the first obstacle; The A* algorithm is used to calculate the shortest path for the drone to pass through the first obstacle.

Images