Unmanned aerial vehicle-based navigation inspection system suitable for multiple types of power towers

Abstract

The invention discloses a navigation inspection system suitable for multiple types of power towers based on an unmanned aerial vehicle, and the system comprises the unmanned aerial vehicle which is used for carrying out the flight and image collection, and carrying out the auxiliary navigation according to a flight control command sent by an edge box; the camera is arranged below the unmanned aerial vehicle and is used for receiving a picture, sent by the unmanned aerial vehicle, of the current tower; the edge box is used for acquiring a shot image of the current tower, carrying out image processing and target identification, calculating a relationship between current position information and a target tower position through an intelligent processing unit therein, if position deviation occurs, carrying out position correction by adopting a correction algorithm, and sending unmanned aerial vehicle correction flight data to the unmanned aerial vehicle; and the RTK orientation module is arranged in the unmanned aerial vehicle and is used for acquiring the current position information of the unmanned aerial vehicle, and the automatic identification and inspection of the electric power tower are realized by combining the open source unmanned aerial vehicle and the edge calculation box.

Description

Technical Field

[0001] The present invention relates to the technical field of navigation inspection of power poles and towers, and in particular to an intelligent inspection system for distribution towers based on unmanned aerial vehicles. Background Art

[0002] Currently, inspections of power transmission systems mostly rely on manual inspections or GPS-based drone inspection methods. These methods generally include waypoint-based inspections, tower inspection-based inspections, and transmission line inspection-based inspections. However, these methods still require accurate tower knowledge in actual operation and may have measurement errors.

[0003] The disadvantage of the prior art is that it relies on prior tower knowledge, which limits the universality and automation of the inspection method. For example, the invention patent with publication number CN111814980B provides an ontology-based transmission line tower knowledge base system and its construction method. The invention integrates and sorts out the transmission line towers, abstracts and uniformly models the transmission line towers, and improves the sharing and reusability of the transmission line tower knowledge in the field of tower inspection. However, the patent still relies on prior tower knowledge.

[0004] In order to solve the above problems, the invention patent with publication number CN118466565A discloses an intelligent inspection system for distribution towers based on drones. However, it can only inspect straight towers, which has great restrictions on the distribution types of poles and towers. In addition, the correction method is too complicated, and the inspection accuracy and efficiency cannot be guaranteed. Summary of the invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the prior art, the present invention provides a distribution tower intelligent inspection system based on drones, which solves the technical problems in the above-mentioned prior art.

[0006] Technical solution: The present invention provides a navigation inspection system based on drones applicable to multiple types of power towers, the system comprising:

[0007] The drone is responsible for flight and image acquisition, and assists in navigation based on the flight control commands sent by the edge box;

[0008] The camera is set under the drone and is used to receive the photos sent by the drone to take pictures of the current tower;

[0009] The edge box is used to obtain the image of the current tower, perform image processing and target recognition, and calculate the relationship between the current position information and the target tower position through the intelligent processing unit. If there is a position deviation, the correction algorithm is used to correct the position, and the correction flight data of the drone is sent to the drone to adjust the heading angle. The above calculation process is repeated until the drone completes the inspection of all target towers;

[0010] The RTK orientation module is set in the drone and is used to obtain the current position information of the drone, so as to determine the distance and direction between its own position and the target tower position, and write the current position information into the picture when the camera takes pictures.

[0011] Further, including:

[0012] The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is the second tower, the processing method used is recorded as the first processing method:

[0013] After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command;

[0014] The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take a photo H1, and then continues to fly vertically upwards;

[0015] The drone flies to the pole top position, i.e. the second target position, and stops to take a photo H2. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0016] The drone continues to fly vertically upwards, and the top of the pole appears at the bottom quarter of the photo, which is point 3. Stop and take photo H3;

[0017] The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result;

[0018] The drone flies vertically upward to the flight altitude, maintains the heading angle of the drone, adjusts the pitch angle of the gimbal, and starts flying toward the second tower perpendicular to the ground.

[0019] The drone continues to fly, with the gimbal camera lens pointing vertically downward. When it identifies the second pole tower, the drone starts to slow down. When the second pole tower reaches a quarter of the distance above the photo, that is, at point 4, the drone stops to take pictures.

[0020] The drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward.

[0021] The drone continues to fly forward. When the tower head is one quarter of the way down in the photo, it stops to take pictures, that is, at point six, and shoots vertically downward. After shooting at each level of the tower, it needs to return the corresponding azimuth and coordinates.

[0022] Further, including:

[0023] During the photography and processing of the first tower and the second tower, the intelligent processing unit also compares and verifies the data information corresponding to the image obtained by the horizontal shooting of the camera with the data information collected by the vertical shooting to confirm whether the drone data collection is normal. If the compared data is within the error range, the aircraft parameters are correct and the operation continues.

[0024] Further, including:

[0025] During the photo-taking and processing of the first tower and the second pole tower, the intelligent processing unit also corrects the drone after six photos are taken, so that the drone always flies above the line, thereby ensuring that the drone can find the next pole tower smoothly.

[0026] Further, including:

[0027] The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is larger than the second tower, the processing method adopted includes:

[0028] First, determine the type of tower from the first base tower to the target tower. If the current tower is a straight tower, continue to use the first processing method to control the drone; otherwise,

[0029] If the current pole tower is a corner tower, the second processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The second processing method includes:

[0030] After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command;

[0031] The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take pictures, and then continues to fly vertically upwards;

[0032] The drone flies to the pole top position, i.e. the second target position, and stops to take a photo. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0033] The drone continues to fly vertically upwards, and the top of the pole appears at a quarter of the bottom of the photo, which is the second target position. The drone stops to take a photo.

[0034] The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result;

[0035] The drone flies forward, with the gimbal camera lens pointing vertically downward. When the drone intelligent processing unit recognizes the current tower head, the drone slows down. When the tower head reaches the fourth target position, the drone stops flying and takes the fourth photo H1. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward.

[0036] The drone continues to fly forward. When the tower head is at the center of the photo, that is, the fifth target position, the drone stops and takes a photo H2.

[0037] The drone continues to fly forward horizontally. When the tower head is one quarter below the photo, the drone stops to take a photo at H3, the sixth target position, and shoots vertically downwards. The drone is corrected at this point.

[0038] The intelligent identification unit identifies the type of the next tower. If it is a straight tower, the first processing method is used. Otherwise, if it is a corner tower, the second processing method is used to inspect.

[0039] Further, including:

[0040] The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is larger than the second tower, the processing method adopted also includes:

[0041] If the current pole tower is a branch tower, a third processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The third processing method includes:

[0042] The intelligent processing unit identifies the current tower type and marks the main line direction and branch line direction. The branch tower position needs to take out two sets of data, namely the front line and the branch line;

[0043] After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command;

[0044] The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take pictures, and then continues to fly vertically upwards;

[0045] The drone flies to the pole top position, i.e. the second target position, and stops to take a photo. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0046] The drone continued to fly vertically upwards, and the top of the pole appeared at the bottom quarter of the photo, that is, point 3, and stopped to take pictures;

[0047] The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result;

[0048] The drone continues to fly, with the gimbal camera lens pointing vertically downward. When it identifies the second tower, the drone starts to slow down. When the current branch tower reaches a quarter of the distance above the photo, the drone stops at point 4 to take photos.

[0049] The drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward.

[0050] The drone continues to fly forward. When the tower head is at one quarter of the bottom of the photo, stop the drone to take a photo, that is, at point 6, and shoot vertically downward. After shooting at each level of the tower, the corresponding azimuth and coordinates must be returned.

[0051] After the sixth photo was taken, the drone retreated. When the head of the branch tower was in the center of the picture, the drone stopped flying backwards and switched to manual intervention in the background. The corresponding main line and branch line data were recorded in the background so that the drone could automatically identify them during the next operation. The drone then continued to fly along the work order wire based on the manual selection results in the background.

[0052] Further, including:

[0053] The correction method comprises:

[0054] After completing the shooting of photos H1, H2 and H3 in the horizontal acquisition process, the pole top and pole bottom in the photos are detected using the intelligent processing unit, and photos H1 and H3 are selected as valid data;

[0055] According to the longitude and latitude coordinates of the drone during the two photos, the longitude and latitude differences Δx and Δy between the drone and the tower, as well as the tower height D, are calculated;

[0056] During the drone collection process, the distance between each tower is fixed, even for corner towers. After the wire angle is identified by the wire recognition model, the distance value between the next tower and the current tower is calculated. Therefore, the longitude and latitude coordinates of the next tower are calculated by adding the distance value to the longitude and latitude coordinates of the current tower.

[0057] Adjust the drone gimbal camera angle to 0° so that the camera can shoot horizontally, and calculate the zoom ratio of the camera lens so that the tower in the picture can be detected by the intelligent processing unit. The lens zoom ratio K is obtained by the formula:

[0058]

[0059] Among them, L h represents the height of the image sensor, u is the object distance, that is, the distance between the camera and the tower, H is the actual height corresponding to the screen where the object distance is located, and f represents the minimum focal length;

[0060] When the intelligent processing unit detects that the pixel coordinates of the tower are not in the center of the picture, it calculates the actual distance error value corresponding to the pixel error; the actual distance error is converted into longitude and latitude coordinates to obtain the corrected longitude and latitude of the target position. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

[0061] Further, including:

[0062] The actual distance error value corresponding to the calculated pixel error is expressed as:

[0063]

[0064] Where Δy and Δx represent the pixel difference between the height and width of the tower pixel frame center from the image center, respectively. y , D x Represents the pixel values ​​of the actual image height and width, L w represents the width of the image sensor, and ΔH and ΔW represent the actual errors corresponding to the pixel errors in the vertical and horizontal directions.

[0065] Further, including:

[0066] The actual distance error is converted into longitude and latitude coordinates:

[0067]

[0068] Among them, M and N represent the current longitude and latitude of the drone; OW represents the horizontal offset of the tower pixel frame from the current position of the drone; a represents the azimuth perpendicular to the flight direction of the drone, which is used to determine whether the tower pixel frame is on the left or right side of the drone; P and Q are the relationship coefficients between the actual distance and the longitude and latitude, respectively, which are calculated based on the current longitude and latitude values; X and Y represent the corrected longitude and latitude of the target position, and the actual error distance of the Z axis can be directly added to the current target height. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

[0069] Further, including:

[0070] The pixel differences between the height and width of the center of the tower pixel frame from the center of the image are expressed as:

[0071]

[0072] Among them, suppose the latitude and longitude coordinates of the drone when taking the two photos are A(A j , A w ),B(B j , B w ); the pixel coordinates of the pole top and pole bottom corresponding to photo H1 are M1(a1, b1), N1(c1, d1), and the pixel coordinates of the pole top and pole bottom corresponding to photo H3 are M2(a2, b2), N2(c2, d2); calculate the moving azimuth angle θ` of the drone:

[0073]

[0074] Calculate the distance AB between A and B:

[0075]

[0076] Where R represents the radius of the earth;

[0077] Calculate the x- and y-components AC and BC of AB:

[0078]

[0079] Calculate the deflection angle β of the M1 coordinate relative to y:

[0080]

[0081] Where a and b are the total length and width of the photo in pixels, and OM1 represents the pixel distance from M1 to the center of the picture;

[0082] Calculate the actual coordinates of M1 coordinates as (h, t):

[0083]

[0084] Further, including:

[0085] After the deviation is corrected, the drone needs to execute the detection command and detect the direction of the wire through the wire recognition model to verify the azimuth of the wire. At the same time, it calls the pole tower recognition model to detect the position of the pole tower and evaluate the ups and downs of the pole tower in the line to confirm whether it deviates from the direction and whether there is a pole tower that is too high or too low. If there is any abnormality, an alarm will be issued.

[0086] Further, including:

[0087] The wire recognition model and the tower recognition model are both obtained by training the yolov5 target detection model.

[0088] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0089] The present invention realizes the automatic identification and inspection of power poles and towers by combining open source drones and edge computing boxes. Using the AI ​​program running on the edge box, frames are extracted from the video stream of the camera carried by the drone for real-time target recognition. Using advanced deep learning algorithms such as yolov5s, it is possible to accurately identify power poles and towers and their locations, and accurately detect the direction of the wires. Through automated flight control, drones can accurately aim at various types of tower locations, take photos and record, calculate longitude and latitude information, and correct the flight of drones based on relevant data information, provide more accurate route guidance, and provide effective geographic information data for subsequent inspections. This technological innovation not only improves inspection efficiency and accuracy, but also reduces labor costs and safety risks, providing important support for intelligent inspections in the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 This is a schematic diagram of the drone image acquisition direction for a straight tower according to Embodiment 1 of the present invention;

[0091] Figure 2 This is a schematic diagram of the drone image acquisition direction for the corner tower according to Example 2 of the present invention;

[0092] Figure 3 This is a schematic diagram of the drone image acquisition direction for the branch tower described in Example 3 of the present invention. DETAILED DESCRIPTION

[0093] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0094] Example 1

[0095] The present invention provides a navigation inspection system based on an unmanned aerial vehicle and applicable to multiple types of power poles and towers, the system comprising:

[0096] The drone is responsible for flight and image acquisition, and assists in navigation based on the flight control commands sent by the edge box;

[0097] The camera is set under the drone and is used to receive the photos sent by the drone to take pictures of the current tower;

[0098] The edge box is used to obtain the image of the current tower, perform image processing and target recognition, and calculate the relationship between the current position information and the target tower position through the intelligent processing unit. If there is a position deviation, the correction algorithm is used to correct the position, and the correction flight data of the drone is sent to the drone to adjust the heading angle. The above calculation process is repeated until the drone completes the inspection of all target towers;

[0099] The RTK orientation module is set in the drone and is used to obtain the current position information of the drone, so as to determine the distance and direction between its own position and the target tower position, and write the current position information into the picture when the camera takes pictures.

[0100] In summary, the software part of this technical solution consists of the embedded software on the edge box and the flight control software on the drone. The embedded software on the edge box is responsible for acquiring images from the camera video stream in real time and performing target recognition through deep learning algorithms (such as yolov5). The recognized power tower location information will be transmitted to the flight control software.

[0101] The flight control software runs on the drone and is responsible for receiving the target location information from the edge box and controlling the drone through the drone's flight control system. Based on the received target location information, the drone will automatically adjust its flight attitude and fly to the target power tower location. Once it reaches the target location, the drone will take photos and record GPS information, and then continue the inspection mission based on the new target location information.

[0102] The entire software system is connected via a 4G network, enabling real-time communication and data transmission between the edge box and the drone. The edge box is responsible for real-time target recognition and data processing, while the drone is responsible for executing flight missions based on the recognition results. This software solution can realize automated inspections of power towers, improving inspection efficiency and accuracy.

[0103] The hardware part of this technical solution includes:

[0104] Drone: Uses an open source multi-rotor drone platform, providing SDK and open interfaces.

[0105] Edge box: An embedded device using Rockchip RK3588 chip, with built-in 4G network module and Ethernet interface, and operating system Ubuntu20.04.

[0106] Camera: Installed under the drone to capture images of power poles and towers.

[0107] RTK module: installed on the drone to obtain flight position information.

[0108] 4G network module: installed on the drone and edge box respectively, used to realize data transmission and communication of cloud center station.

[0109] The drone and edge box communicate through the internal IP network to transmit image data and control instructions. The edge box connects to the camera to obtain images of power poles in real time. The drone has a built-in RTK module to obtain current location information and write image EXIF ​​metadata when taking pictures.

[0110] Among them, the drone is responsible for flight and image acquisition, and performs AI-assisted navigation based on the flight control commands sent by the edge box and the target location information. The edge box is responsible for image processing and target recognition, using deep learning algorithms to identify power towers and calculate the target tower location information. The RTK module provides current location information to help the edge box's positioning algorithm determine the distance and direction between its own position and the target tower position.

[0111] Furthermore, the above-mentioned image processing and target recognition are mainly image acquisition operations, which mainly include the following steps:

[0112] The intelligent processing unit therein calculates the relationship between the current position information and the position of the target tower, including: when the target tower is the second tower, that is, the collected towers include only two, the processing method used is recorded as the first processing method, such as Figure 1 As shown:

[0113] After receiving the task work order, the S11 drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image and gives the drone controller a take-off command after identifying the first tower.

[0114] The S12 drone flies vertically upwards. When the camera recognizes the entire tower, it stops at the first target position of the first tower to take a photo H1, and then continues to fly vertically upwards;

[0115] S13 The drone flies to the pole top position, i.e. the second target position, and stops to take a photo H2. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0116] The S14 drone continued to fly vertically upwards, and the top of the pole appeared at the bottom quarter of the photo, that is, point 3. The drone stopped to take a photo H3;

[0117] The S15 UAV uses the first target position and the third target position to take photos to calculate the pole top elevation. The intelligent processing unit sends the calculation results to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation results;

[0118] The S15 drone flies vertically upward to the flight altitude, maintains the heading angle of the drone, adjusts the pitch angle of the gimbal, and starts flying toward the second tower perpendicular to the ground.

[0119] The S15 drone continued to fly, with the gimbal camera lens pointing vertically downward. When the AI ​​identified the second pole tower, Tower 2, the drone began to slow down. When Tower 2 reached a quarter of the distance above the photo, i.e., point 4, the drone stopped to take a photo H4;

[0120] The S15 drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo H5. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90° and the photo is taken vertically downward.

[0121] The S15 drone continues to fly forward. When the tower head is one quarter below the photo, stop and take a photo at H6, which is point 6, and shoot vertically downward. After shooting at each level of the tower, the corresponding azimuth and coordinates must be returned.

[0122] When arriving at position 6, the mission system autonomously hovers to take photos and transmits them back until it receives confirmation from the intelligent unit before continuing the mission.

[0123] Furthermore, in this embodiment, during the photographing and data collection process of the first tower and the second tower, the intelligent processing unit further includes:

[0124] S151 compares and verifies the data information corresponding to the image obtained by the horizontal shooting of the camera with the data information collected by the vertical shooting to confirm whether the drone data collection is normal. If the compared data is within the error range, the aircraft parameters are correct and the operation continues.

[0125] During the process of photographing and processing the first tower and the second pole tower by S152, the intelligent processing unit also corrects the deviation of the drone after six photos are taken, so that the drone always flies above the line, thereby ensuring that the drone can find the next pole tower smoothly.

[0126] After the horizontal shooting of the S153 first tower is completed, the gimbal camera angle needs to be adjusted to -90°, the wire direction needs to be identified based on the wire identification model, and the heading angle needs to be corrected using the wire direction.

[0127] S154 adjusts the gimbal camera angle to 0° so that the camera can shoot horizontally for subsequent tower identification. It then calls the tower identification model to detect the tower. The intelligent processing unit outputs the horizontal offset or coordinates to the drone control unit based on the identification results, controls the movement of the aircraft, and ensures that the tower is in the center of the picture, thereby achieving deviation correction.

[0128] The S16 intelligent processing unit corrects the drone's flight, adjusts the drone so that it flies above the line, adjusts the heading angle, and then accelerates to fly along the line.

[0129] In this embodiment, the above-mentioned deviation correction operation is implemented by the following scheme:

[0130] Step 1: After completing the shooting of photos H1, H2 and H3 in the horizontal acquisition process, the pole top and pole bottom in the photos are detected by using the intelligent processing unit, and photos H1 and H3 are selected as valid data;

[0131] Step 2: Calculate the latitude and longitude differences Δx and Δy between the drone and the tower, as well as the tower height D, based on the latitude and longitude coordinates of the drone during the two photos.

[0132] Step 3: During the drone collection process, the distance between each tower is fixed, even for corner towers. After the wire angle is identified by the wire recognition model, the distance value between the next tower and the current tower is calculated. Therefore, the longitude and latitude coordinates of the next tower are calculated by adding the distance value to the longitude and latitude coordinates of the current tower.

[0133] Step 4: Adjust the camera angle of the drone gimbal to 0° so that the camera can shoot horizontally, and calculate the zoom ratio of the camera lens so that the tower in the picture can be detected by the intelligent processing unit. The lens zoom ratio K is obtained by the formula:

[0134]

[0135] Among them, L h represents the height of the image sensor, u is the object distance, that is, the distance between the camera and the tower, H is the actual height corresponding to the screen where the object distance is located, and f represents the minimum focal length;

[0136] Step 5: When the intelligent processing unit detects that the pixel coordinates of the tower are not in the center of the picture, the actual distance error value corresponding to the pixel error is calculated; the actual distance error is converted into longitude and latitude coordinates to obtain the corrected longitude and latitude of the target position. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

[0137] Furthermore, in this embodiment, the actual distance error value corresponding to the above-mentioned pixel error is expressed as:

[0138]

[0139] Where Δy and Δx represent the pixel difference between the height and width of the tower pixel frame center from the image center, respectively. y , D x Represents the pixel values ​​of the actual image height and width, L w represents the width of the image sensor, and ΔH and ΔW represent the actual errors corresponding to the pixel errors in the vertical and horizontal directions.

[0140] In this embodiment, the actual distance error is converted into longitude and latitude coordinates:

[0141]

[0142] Among them, M and N represent the current longitude and latitude of the drone; OW represents the horizontal offset of the tower pixel frame from the current position of the drone; a represents the azimuth perpendicular to the flight direction of the drone, which is used to determine whether the tower pixel frame is on the left or right side of the drone; P and Q are the relationship coefficients between the actual distance and the longitude and latitude, respectively, which are calculated based on the current longitude and latitude values; X and Y represent the corrected longitude and latitude of the target position, and the actual error distance of the Z axis can be directly added to the current target height. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

[0143] In this embodiment, the pixel difference between the height and width of the center of the tower pixel frame from the center of the image is expressed as:

[0144]

[0145] Among them, suppose the latitude and longitude coordinates of the drone when taking the two photos are A(A j , A w ), B(Bj , B w ); the pixel coordinates of the pole top and pole bottom corresponding to photo H1 are M1(a1, b1), N1(c1, d1), and the pixel coordinates of the pole top and pole bottom corresponding to photo H3 are M2(a2, b2), N2(c2, d2); calculate the moving azimuth angle θ` of the drone:

[0146]

[0147] Calculate the distance AB between A and B:

[0148]

[0149] Where R represents the radius of the earth;

[0150] Calculate the x- and y-components AC and BC of AB:

[0151]

[0152] Calculate the deflection angle β of the M1 coordinate relative to y:

[0153]

[0154] Where a and b are the total length and width of the photo in pixels, and OM1 represents the pixel distance from M1 to the center of the picture;

[0155] Calculate the actual coordinates of M1 coordinates as (h, t):

[0156]

[0157] The inspection system of this embodiment also includes:

[0158] After the deviation is corrected, the drone needs to execute the detection command and detect the direction of the wire through the wire recognition model to verify the azimuth of the wire. At the same time, it calls the pole tower recognition model to detect the position of the pole tower and evaluate the ups and downs of the pole tower in the line to confirm whether it deviates from the direction and whether there is a pole tower that is too high or too low. If there is any abnormality, an alarm will be issued.

[0159] Among them, the wire recognition model and the pole tower recognition model are both obtained by training the yolov5 target detection model. In this embodiment, the edge box is responsible for real-time image processing and target recognition of the power pole tower, and a deep learning algorithm (yolov5) is used.

[0160] Example 2

[0161] In this embodiment, the intelligent processing unit of the present invention calculates the relationship between the current position information and the position of the target tower, including: when the target tower is larger than the second tower, that is, the number of towers exceeds two, the processing method used includes:

[0162] First, determine the type of tower from the first base tower to the target tower. If the current tower is a straight tower, continue to use the first processing method to control the drone; otherwise,

[0163] like Figure 2 As shown, if the current pole tower is a corner tower, the second processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The second processing method includes firstly using a similar method in the first processing method for the first tower, and then continuing to fly horizontally along the line after processing. If it is identified that the nearest pole tower in front is a corner tower, a different processing method is used, specifically including:

[0164] Among them, the similar method of the first processing method includes:

[0165] After receiving the task work order, the S21 drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image and gives the drone controller a take-off command after identifying the first tower.

[0166] The S22 drone flies vertically upwards. When the camera recognizes the entire tower, it stops at the first target position of the first tower to take photos, and then continues to fly vertically upwards.

[0167] The S23 drone flies to the pole top position, i.e. the second target position, and stops to take photos. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0168] The S24 drone continued to fly vertically upwards, and the top of the pole appeared in the lower quarter of the photo, which was the second target position. The drone stopped to take pictures.

[0169] The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result;

[0170] For the image acquisition of the corner tower, the following steps are used:

[0171] The S25 drone flies forward, with the gimbal camera lens pointing vertically downward. When the drone intelligent processing unit recognizes the current tower head, the drone slows down. When the tower head reaches the fourth target position, it stops flying and takes the fourth photo H1. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward.

[0172] The S26 drone continues to fly forward. When the tower head is at the center of the photo, i.e. the fifth target position, the drone stops and takes a photo H2.

[0173] The S27 UAV continued to fly forward horizontally. When the tower head was located one quarter below the photo, it stopped to take a photo at H3, the sixth target position, and shot vertically downward, where the UAV was corrected.

[0174] In the present embodiment, H1, H2 and H3 in steps S25 to S27 are for completing the deviation correction operation and marking the photographing point.

[0175] S28 The intelligent identification unit identifies the type of the next pole tower. If it is a straight tower, the first processing method is used. Otherwise, if it is a corner tower, the second processing method is continued to be used for collection until the target pole tower is inspected.

[0176] Furthermore, the inspection system of this embodiment also includes:

[0177] After the deviation is corrected, the drone needs to execute the detection command and detect the direction of the wire through the wire recognition model to verify the azimuth of the wire. At the same time, it calls the pole tower recognition model to detect the position of the pole tower and evaluate the ups and downs of the pole tower in the line to confirm whether it deviates from the direction and whether there is a pole tower that is too high or too low. If there is any abnormality, an alarm will be issued.

[0178] Among them, the wire recognition model and the pole tower recognition model are both obtained by training the yolov5 target detection model. In this embodiment, the edge box is responsible for real-time image processing and target recognition of the power pole tower, and a deep learning algorithm (yolov5) is used.

[0179] The correction operation adopted in this embodiment is:

[0180] Step 1: Calculate the latitude, longitude and height of the base tower;

[0181] After completing the shooting of photos H1, H2 and H3 in the horizontal acquisition process, the intelligent processing unit is used to detect the pole top and pole bottom in the photos, and photos H1 and H3 are selected as valid data.

[0182] Assume that the latitude and longitude coordinates of the drone when taking the two photos are A(A j , A w ), B(B j , B w ); the pixel coordinates of the pole top and pole bottom corresponding to photo H1 are M1(a1, b1), N1(c1, d1), and the pixel coordinates of the pole top and pole bottom corresponding to photo H3 are M2(a2, b2), N2(c2, d2).

[0183] Calculate the UAV moving azimuth angle θ`:

[0184]

[0185] Calculate the distance AB between A and B:

[0186]

[0187] Where R is the radius of the Earth.

[0188] Calculate the x- and y-components AC and BC of AB:

[0189]

[0190] Calculate the deflection angle β of the M1 coordinate relative to y:

[0191]

[0192] Where a and b are the total length and width of the photo in pixels, and OM1 represents the pixel distance from M1 to the center of the picture.

[0193] Calculate the actual coordinates of M1 coordinates as (h, t):

[0194]

[0195] Calculate the difference in longitude and latitude between the drone and the tower:

[0196]

[0197] Calculate the tower height D:

[0198]

[0199] Finally, the tower latitude and longitude height can be calculated through 1 as (A j +Δx,A w +Δy, D);

[0200] Step 2: During the distribution network collection process, the distance between each base tower is fixed. Even for corner towers, after identifying the conductor angle through the conductor identification model, the distance between the next base tower and the current base tower can still be calculated. Therefore, the longitude and latitude coordinates of the next base tower can be calculated by adding the distance value to the longitude and latitude coordinates of the current base tower.

[0201] Step 3: Adjust the camera angle of the drone gimbal to 0° so that the camera can shoot horizontally, and calculate the zoom ratio of the camera lens so that the tower in the picture can be detected by the intelligent recognition unit. The lens zoom ratio K can be obtained by the formula:

[0202]

[0203] Among them, L hrepresents the height of the image sensor, u is the object distance (the distance between the camera and the tower), H is the actual height corresponding to the screen where the object distance is located, and f represents the minimum focal length.

[0204] Step 4: When the intelligent processing unit detects that the pixel coordinates of the tower are not in the center of the picture, the actual error value corresponding to the pixel error can be calculated by the following formula;

[0205]

[0206] Where Δy and Δx represent the pixel difference between the height and width of the tower pixel frame center from the image center, respectively. y , D x Represents the pixel values ​​of the actual image height and width, L w represents the width of the image sensor, ΔH and ΔW represent the actual errors corresponding to the pixel errors in the vertical and horizontal directions;

[0207] Step 5: Convert the actual distance error into longitude and latitude coordinates;

[0208]

[0209] Among them, M and N represent the current longitude and latitude of the drone; OW represents the horizontal offset of the tower pixel frame from the current position of the drone; a represents the direction perpendicular to the flight direction of the drone, which is used to determine whether the tower pixel frame is on the left or right side of the drone; P and Q are the relationship coefficients between the actual distance and the longitude and latitude, respectively, which are calculated based on the current longitude and latitude values; X and Y represent the corrected longitude and latitude of the target position. The actual error distance of the Z axis can be directly added to the current target height. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

[0210] Example 3

[0211] In this embodiment, image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit, including: when the target tower is larger than the second tower, the processing method adopted also includes:

[0212] like Figure 3 As shown, if the current pole tower is a branch tower, the third processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The third processing method first uses a similar method in the first processing method for the first tower, and then continues to fly horizontally along the line. If the nearest pole tower in front is identified as a branch tower, different processing methods are used, including:

[0213] The S31 intelligent processing unit identifies the current tower type and marks the main line direction and branch line direction. Two sets of data need to be taken out at the branch tower location, namely the front line and the branch line;

[0214] After receiving the task work order, the S32 drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image and gives the drone controller a take-off command after identifying the first tower.

[0215] The S33 drone flies vertically upwards. When the camera recognizes the entire tower, it stops at the first target position of the first tower to take photos, and then continues to fly vertically upwards.

[0216] The S34 drone flies to the pole top position, i.e. the second target position, and stops to take photos. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo.

[0217] The S35 drone continued to fly vertically upwards, and the top of the pole appeared at the bottom quarter of the photo, that is, point 3, and stopped to take pictures;

[0218] The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result;

[0219] The S36 drone continued to fly, with the gimbal camera lens pointing vertically downward. When it identified the second tower, the drone began to slow down. When the current branch tower reached a quarter of the distance above the photo, the drone stopped at point 4 to take photos.

[0220] The S37 drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward.

[0221] The S38 drone continues to fly forward. When the tower head is at one quarter of the bottom of the photo, stop the drone to take photos, that is, at point 6, and shoot vertically downward. After completing the shooting of each level of the tower, the corresponding azimuth and coordinates must be returned.

[0222] S39 After the sixth photo is taken, the drone retreats. When the head of the branch tower is in the center of the picture, the drone stops flying backwards and switches to background manual intervention. The corresponding main line and branch line data are recorded in the background so that the drone can automatically identify them during the next operation. Then the drone continues to fly along the work order wire according to the background manual selection results.

[0223] Therefore, in summary, the edge box of the present invention can only obtain the GPS location information of the current drone and the position of the tower in the photo from the photo, and use multiple photos and the displacement information of the drone to calculate the GPS information, direction angle and distance of the target tower, and ensure high accuracy and low error. The drone needs to perform autonomous navigation and perform flight control tasks based on the received target position information.

[0224] 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.

[0225] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments 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 navigation inspection system based on drones suitable for multiple types of power towers, characterized in that: The system includes: The drone is responsible for flight and image acquisition, and assists in navigation based on the flight control commands sent by the edge box; The camera is set under the drone and is used to receive the photos sent by the drone to take pictures of the current tower; The edge box is used to obtain the image of the current tower, perform image processing and target recognition, and calculate the relationship between the current position information and the target tower position through the intelligent processing unit. If there is a position deviation, the correction algorithm is used to correct the position, and the correction flight data of the drone is sent to the drone to adjust the heading angle. The above calculation process is repeated until the drone completes the inspection of all target towers; The RTK orientation module is set in the drone and is used to obtain the current position information of the drone, so as to determine the distance and direction between its own position and the target tower position, and write the current position information into the picture when the camera takes pictures.

2. The navigation inspection system based on drones and applicable to multiple types of power towers according to claim 1 is characterized in that: The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is the second tower, the processing method used is recorded as the first processing method: After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command; The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take a photo H1, and then continues to fly vertically upwards; The drone flies to the pole top position, i.e. the second target position, and stops to take a photo H2. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo. The drone continues to fly vertically upwards, and the top of the pole appears at the bottom quarter of the photo, which is point 3. Stop and take photo H3; The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result; The drone flies vertically upward to the flight altitude, maintains the heading angle of the drone, adjusts the pitch angle of the gimbal, and starts flying toward the second tower perpendicular to the ground. The drone continues to fly, with the gimbal camera lens pointing vertically downward. When it identifies the second pole tower, the drone starts to slow down. When the second pole tower reaches a quarter of the distance above the photo, that is, at point 4, the drone stops to take pictures. The drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward. The drone continues to fly forward. When the tower head is one quarter of the way down in the photo, it stops to take pictures, that is, at point six, and shoots vertically downward. After shooting at each level of the tower, it needs to return the corresponding azimuth and coordinates.

3. The navigation inspection system based on drones and applicable to various types of power towers according to claim 2 is characterized in that: During the photography and processing of the first tower and the second tower, the intelligent processing unit also compares and verifies the data information corresponding to the image obtained by the horizontal shooting of the camera with the data information collected by the vertical shooting to confirm whether the drone data collection is normal. If the compared data is within the error range, the aircraft parameters are correct and the operation continues.

4. The navigation inspection system based on drones and applicable to various types of power towers according to claim 3 is characterized in that: During the photo-taking and processing of the first tower and the second pole tower, the intelligent processing unit also corrects the drone after six photos are taken, so that the drone always flies above the line, thereby ensuring that the drone can find the next pole tower smoothly.

5. The navigation inspection system based on drones and applicable to multiple types of power towers according to claim 2 is characterized in that: The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is larger than the second tower, the processing method adopted includes: First, determine the type of tower from the first base tower to the target tower. If the current tower is a straight tower, continue to use the first processing method to control the drone; otherwise, If the current pole tower is a corner tower, the second processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The second processing method includes: After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command; The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take pictures, and then continues to fly vertically upwards; The drone flies to the pole top position, i.e. the second target position, and stops to take a photo. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo. The drone continues to fly vertically upwards, and the top of the pole appears at a quarter of the bottom of the photo, which is the second target position. The drone stops to take a photo. The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result; The drone flies forward, with the gimbal camera lens pointing vertically downward. When the drone intelligent processing unit recognizes the current tower head, the drone slows down. When the tower head reaches the fourth target position, the drone stops flying and takes the fourth photo H1. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward. The drone continues to fly forward. When the tower head is at the center of the photo, that is, the fifth target position, the drone stops and takes a photo H2. The drone continues to fly forward horizontally. When the tower head is one quarter below the photo, the drone stops to take a photo at H3, the sixth target position, and shoots vertically downwards. The drone is corrected at this point. The intelligent identification unit identifies the type of the next tower. If it is a straight tower, the first processing method is used. Otherwise, if it is a corner tower, the second processing method is used to inspect.

6. The navigation inspection system based on drones and applicable to various types of power poles and towers according to claim 5 is characterized in that: The image processing and target recognition are performed, and the relationship between the current position information and the position of the target tower is calculated by the intelligent processing unit therein, including: when the target tower is larger than the second tower, the processing method adopted also includes: If the current pole tower is a branch tower, a third processing method is used to enable the drone to find the corresponding target pole tower from the first tower. The third processing method includes: The intelligent processing unit identifies the current tower type and marks the main line direction and branch line direction. The branch tower position needs to take out two sets of data, namely the front line and the branch line; After receiving the task work order, the drone completes the self-check and flight preparation, and starts the intelligent processing unit, which recognizes the image. After identifying the first tower, it gives the drone controller a take-off command; The drone flies vertically upwards, and when the camera recognizes the entire tower, it stops at the first target position of the first tower to take pictures, and then continues to fly vertically upwards; The drone flies to the pole top position, i.e. the second target position, and stops to take a photo. This photo is used to verify the pole top elevation. The second target position is when the pole top appears in the center of the photo. The drone continued to fly vertically upwards, and the top of the pole appeared at the bottom quarter of the photo, that is, point 3, and stopped to take pictures; The drone uses the first target position and the third target position to take photos to calculate the pole top elevation, and the intelligent processing unit sends the calculation result to the flight controller, and the flight controller determines the flight altitude and elevation according to the work order requirements and the calculation result; The drone continues to fly, with the gimbal camera lens pointing vertically downward. When it identifies the second tower, the drone starts to slow down. When the current branch tower reaches a quarter of the distance above the photo, the drone stops at point 4 to take photos. The drone continues to fly forward. When the pole tower is at the center of the photo, that is, point 5, the drone stops to take a photo. This photo is used to obtain the latitude and longitude information of the drone at this point, which is used to verify the latitude and longitude of the pole tower. At this time, the gimbal camera angle is adjusted to -90°, and the photo is taken vertically downward. The drone continues to fly forward. When the tower head is at one quarter of the bottom of the photo, stop the drone to take a photo, that is, at point 6, and shoot vertically downward. After shooting at each level of the tower, the corresponding azimuth and coordinates must be returned. After the sixth photo was taken, the drone retreated. When the head of the branch tower was in the center of the picture, the drone stopped flying backwards and switched to manual intervention in the background. The corresponding main line and branch line data were recorded in the background so that the drone could automatically identify them during the next operation. The drone then continued to fly along the work order wire based on the manual selection results in the background.

7. The navigation inspection system based on drones and applicable to various types of power towers according to claim 4 is characterized in that: The correction method comprises: After completing the shooting of photos H1, H2 and H3 in the horizontal acquisition process, the pole top and pole bottom in the photos are detected using the intelligent processing unit, and photos H1 and H3 are selected as valid data; According to the longitude and latitude coordinates of the drone during the two photos, the longitude and latitude differences Δx and Δy between the drone and the tower, as well as the tower height D, are calculated; During the drone collection process, the distance between each tower is fixed, even for corner towers. After the wire angle is identified by the wire recognition model, the distance value between the next tower and the current tower is calculated. Therefore, the longitude and latitude coordinates of the next tower are calculated by adding the distance value to the longitude and latitude coordinates of the current tower. Adjust the drone gimbal camera angle to 0° so that the camera can shoot horizontally, and calculate the zoom ratio of the camera lens so that the tower in the picture can be detected by the intelligent processing unit. The lens zoom ratio K is obtained by the formula: Among them, L h represents the height of the image sensor, u is the object distance, that is, the distance between the camera and the tower, H is the actual height corresponding to the screen where the object distance is located, and f represents the minimum focal length; When the intelligent processing unit detects that the pixel coordinates of the tower are not in the center of the picture, it calculates the actual distance error value corresponding to the pixel error; the actual distance error is converted into longitude and latitude coordinates to obtain the corrected longitude and latitude of the target position. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

8. The navigation inspection system based on drones and applicable to various types of power poles and towers according to claim 7 is characterized in that: The actual distance error value corresponding to the calculated pixel error is expressed as: Where Δy and Δx represent the pixel difference between the height and width of the tower pixel frame center from the image center, respectively. y , D x Represents the pixel values ​​of the actual image height and width, L w represents the width of the image sensor, and ΔH and ΔW represent the actual errors corresponding to the pixel errors in the vertical and horizontal directions.

9. The navigation inspection system based on drones and applicable to various types of power poles and towers according to claim 8, characterized in that: The actual distance error is converted into longitude and latitude coordinates: Among them, M and N represent the current longitude and latitude of the drone; OW represents the horizontal offset of the tower pixel frame from the current position of the drone; a represents the azimuth perpendicular to the flight direction of the drone, which is used to determine whether the tower pixel frame is on the left or right side of the drone; P and Q are the relationship coefficients between the actual distance and the longitude and latitude, respectively, which are calculated based on the current longitude and latitude values; X and Y represent the corrected longitude and latitude of the target position, and the actual error distance of the Z axis can be directly added to the current target height. Finally, the drone flies to the corrected longitude and latitude coordinates to complete the correction.

10. The navigation inspection system based on drones and applicable to various types of power towers according to claim 9, characterized in that: The pixel differences between the height and width of the center of the tower pixel frame from the center of the image are expressed as: Among them, suppose the latitude and longitude coordinates of the drone when taking the two photos are A(A i , A w ), B(B j , B w ); the pixel coordinates of the pole top and pole bottom corresponding to photo H1 are M1(a1, b1), N1(c1, d1), and the pixel coordinates of the pole top and pole bottom corresponding to photo H3 are M2(a2, b2), N2(c2, d2); calculate the moving azimuth angle θ` of the drone: Calculate the distance AB between A and B: Where R represents the radius of the earth; Calculate the x- and y-components AC and BC of AB: Calculate the deflection angle β of the M1 coordinate relative to y: Where a and b are the total length and width of the photo in pixels, and OM1 represents the pixel distance from M1 to the center of the picture; Calculate the actual coordinates of M1 coordinates as (h, t):

11. The navigation inspection system based on drones and applicable to various types of power poles and towers according to claim 10, characterized in that: After the deviation is corrected, the drone needs to execute the detection command and detect the direction of the wire through the wire recognition model to verify the azimuth of the wire. At the same time, it calls the pole tower recognition model to detect the position of the pole tower and evaluate the ups and downs of the pole tower in the line to confirm whether it deviates from the direction and whether there is a pole tower that is too high or too low. If there is any abnormality, an alarm will be issued.

12. The navigation inspection system based on drones and applicable to multiple types of power poles and towers according to claim 11, characterized in that: The wire recognition model and the tower recognition model are both obtained by training the yolov5 target detection model.

Images