AI vision-based distribution network unmanned aerial vehicle ground effect flight method, device and medium

Abstract

The application discloses a kind of based on AI vision's network unmanned plane ground effect flight method, equipment and medium, the method includes: in the height calibration stage before unmanned plane flies to next target tower pole, first real-time image in flight direction is collected in real time;According to the image width proportion of distribution wire in first real-time image, dynamically adjust the height of unmanned plane, until reach target tower safety height;After unmanned plane flies over target tower pole according to target safety height, second real-time image in flight direction is collected in real time by vision sensor;According to the image width proportion of distribution wire in each second real-time image and the real-time positioning information of unmanned plane, control unmanned plane to target tower as starting point, follow the height fluctuation of distribution wire in flight direction and carry out flight height change, the technical scheme of the embodiment of the application realizes unmanned plane autonomous height adjustment by visual intelligent analysis, greatly reduces equipment cost and operation complexity, significantly improves the efficiency and quality of unmanned plane inspection.

Description

Technical Field

[0001] This invention relates to the field of drone technology, and in particular to a method, equipment, and medium for ground-following flight of a power distribution drone based on AI vision. Background Technology

[0002] With the rapid advancement of smart grid construction, the demand for automated and intelligent inspection of distribution network lines is becoming increasingly urgent. However, distribution network lines are mostly located in areas with complex terrain and significant height differences between adjacent towers, which places extremely high demands on the terrain-following flight capabilities of drones.

[0003] Current technologies primarily employ terrain simulation based on DEM (Digital Elevation Model) or DSM (Digital Surface Model) data, along with onboard LiDAR-equipped flight operations. However, these technologies require pre-downloading of terrain data, resulting in low accuracy. Furthermore, they necessitate the use of LiDAR-equipped UAVs for point cloud scanning and computational calculations to achieve onboard terrain simulation, making it difficult to guarantee flight safety and inspection quality for power distribution network inspections. Summary of the Invention

[0004] This invention provides a method, device, and medium for ground-following flight of power distribution network drones based on AI vision. This enables drones to autonomously follow the terrain undulations of power poles for ground-following flight without relying on lidar equipment or pre-installing terrain data.

[0005] According to one aspect of the present invention, a method for ground-following flight of a power distribution drone based on AI vision is provided, the method comprising:

[0006] During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time by the visual sensor mounted on the drone gimbal.

[0007] Based on the proportion of the image width of the power distribution wires in each first real-time image, the flight altitude of the UAV is dynamically adjusted until the UAV reaches the target safe altitude that matches the target tower.

[0008] After the drone flies past the target tower at the target safe altitude, it continues to collect a second real-time image of the flight direction through the visual sensor.

[0009] Based on the proportion of the power distribution line's width in each second real-time image and the UAV's real-time positioning information, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution line in the flight direction.

[0010] According to another aspect of the present invention, a ground-following flight device for power distribution drones based on AI vision is provided, the device comprising:

[0011] The altitude calibration module is used to collect the first real-time image of the flight direction in real time through the visual sensor mounted on the drone gimbal during the altitude calibration phase before the drone flies to the next target tower.

[0012] The dynamic altitude adjustment module is used to dynamically adjust the drone's flight altitude based on the image width ratio of the power distribution wires in each first real-time image until the drone reaches the target safe altitude that matches the target tower.

[0013] The target tower transition module is used to continue to acquire a second real-time image of the flight direction through a visual sensor after the UAV has flown past the target tower at the target safe altitude.

[0014] The follow-flight module is used to control the UAV to change its flight altitude by following the height fluctuations of the power distribution wires in the flight direction, based on the image width ratio of the power distribution wires in each second real-time image and the UAV's real-time positioning information.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] At least one processor; and

[0017] A memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform a method for ground-following flight of a power distribution drone based on AI vision, as described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement a method for ground-following flight of a power distribution drone based on AI vision as described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the steps of the method as described in any embodiment of the present invention.

[0021] The technical solution of this invention involves a height calibration phase before the drone flies to the next target tower. A first real-time image along the flight path is acquired in real-time using a visual sensor mounted on the drone's gimbal. The drone's flight altitude is dynamically adjusted based on the width ratio of the power distribution wires in each of the first real-time images until a target safe altitude matching the target tower is reached. After the drone flies past the target tower at the target safe altitude, a second real-time image along the flight path is acquired in real-time using the visual sensor. Based on the width ratio of the power distribution wires in each of the second real-time images and the drone's real-time positioning information, the drone is controlled to change its flight altitude from the target tower, following the height fluctuations of the power distribution wires along the flight path, thus achieving autonomous terrain-following flight. This novel AI vision-based terrain-following flight method for power distribution drones acquires wire images in real-time using a visual sensor and analyzes the width ratio, enabling the drone to accurately follow the spatial direction of the wires without relying on lidar or pre-installed terrain data, significantly improving the safety and economy of power grid inspection.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a method for ground-following flight of a power distribution drone based on AI vision, according to Embodiment 1 of the present invention.

[0025] Figure 2 This is a flowchart of another AI vision-based method for the terrain-following flight of a power distribution drone according to Embodiment 2 of the present invention;

[0026] Figure 3 This is a flowchart of another AI vision-based method for the terrain-following flight of a power distribution drone according to Embodiment 3 of the present invention;

[0027] Figure 4 This is a schematic diagram of the structure of a power distribution drone based on AI vision and following the terrain, according to Embodiment 4 of the present invention.

[0028] Figure 5 This is a schematic diagram of the structure of an electronic device that implements a method for the ground-following flight of a power distribution drone based on AI vision, according to an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] Example 1

[0032] Figure 1 This is a flowchart of a method for a distribution network drone to follow terrain based on AI vision, provided in Embodiment 1 of the present invention. This embodiment is applicable to the autonomous inspection operation of distribution network lines under complex terrain conditions by drones. The method can be executed by a distribution network drone following terrain based on AI vision. The device can be implemented in hardware and / or software and is generally configured in electronic devices.

[0033] Correspondingly, such as Figure 1 As shown, the method includes:

[0034] S110. During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time through the visual sensor mounted on the drone gimbal.

[0035] The altitude calibration phase can be understood as the preprocessing stage of the UAV's terrain-following flight control, establishing an initial flight baseline through visual sensors mounted on the gimbal. This phase occurs during the transition period when the UAV is about to leave the current pole and fly to the next target pole, mainly completing the initialization of control system parameters and confirmation of a safe flight altitude. The first real-time image can be understood as the visual data collected during the altitude calibration phase, containing clear image features of the connecting wires of the target pole.

[0036] In this embodiment, the UAV first performs an initialization process before flying towards the new tower, acquiring real-time image data of the line ahead through the gimbal vision device. This process can use a pre-trained target recognition model, such as the improved YOLOv11 framework, to identify and locate the conductor, and calculate the proportion of the conductor's width in the image (i.e., the first real-time image) using radio=w / W, where w is the detected conductor pixel width and W is the image frame width.

[0037] S120. Based on the proportion of the power distribution wires in each first real-time image, dynamically adjust the drone's flight altitude until the drone reaches the target safe altitude that matches the target tower.

[0038] The target safe altitude can be understood as the optimal flight altitude at which a drone can ensure inspection quality while avoiding collision risks when flying over a specific pole. This altitude is dynamically determined by the width ratio of the conductor in the image. When the image width ratio radio value is stable within the optimal range, it indicates that the drone is in an ideal position that can clearly capture the details of the conductor while maintaining a safe distance.

[0039] In this embodiment, the drone's flight altitude is dynamically adjusted based on the calculated radio value of the conductor in the first real-time image. When the radio value is less than a preset descent threshold (e.g., 15%), the drone performs a constant-speed descent operation. When the radio value is greater than a preset ascent threshold (e.g., 20%), the drone performs a constant-speed ascent operation. This process continues until the drone is within the optimal width ratio range, thus completing the dynamic altitude adjustment operation of the altitude calibration phase.

[0040] S130: After the UAV flies past the target tower at the target safe altitude, it continues to collect a second real-time image in the flight direction through the visual sensor.

[0041] The second real-time image can be understood as flight direction image data continuously collected by the visual sensor after the UAV completes altitude calibration and flies over the target tower. This type of image records the real-time spatial distribution characteristics of the guide wire after the UAV enters cruise flight mode, providing a dynamic adjustment basis for subsequent terrain-following flight.

[0042] S140. Based on the image width ratio of the power distribution wire in each second real-time image and the real-time positioning information of the UAV, control the UAV to take the target tower as the starting point and follow the height fluctuation of the power distribution wire in the flight direction to change the flight altitude.

[0043] In this embodiment, the UAV achieves precise guide wire following by fusing visual data and positioning information. This method utilizes real-time acquired second-real-time images to analyze the width ratio of the guide wire in the frame, and combines this with position information provided by a satellite positioning system to calculate the spatial relationship between the UAV and the guide wire. An intelligent algorithm dynamically adjusts the flight altitude, enabling the UAV to autonomously follow the guide wire's changing trajectory in three-dimensional space.

[0044] The technical solution of this invention involves a height calibration phase before the drone flies to the next target tower. A first real-time image along the flight path is acquired in real-time using a visual sensor mounted on the drone's gimbal. The drone's flight altitude is dynamically adjusted based on the width ratio of the power distribution wires in each of the first real-time images until a target safe altitude matching the target tower is reached. After the drone flies past the target tower at the target safe altitude, a second real-time image along the flight path is acquired in real-time using the visual sensor. Based on the width ratio of the power distribution wires in each of the second real-time images and the drone's real-time positioning information, the drone is controlled to change its flight altitude from the target tower, following the height fluctuations of the power distribution wires along the flight path, thus achieving autonomous terrain-following flight. This novel AI vision-based terrain-following flight method for power distribution drones acquires wire images in real-time using a visual sensor and analyzes the width ratio, enabling the drone to accurately follow the spatial direction of the wires without relying on lidar or pre-installed terrain data, significantly improving the safety and economy of power grid inspection.

[0045] Example 2

[0046] Figure 2 This is a flowchart of another AI vision-based method for the ground-following flight of a power distribution drone, provided in Embodiment 2 of the present invention. This embodiment is based on and optimized from the above embodiments. Specifically, the operation of "dynamically adjusting the drone's flight altitude according to the image width ratio of the power distribution conductors in each first real-time image until the drone reaches the target safe altitude matching the target tower" has been refined.

[0047] Correspondingly, such as Figure 2 As shown, the method includes:

[0048] S210. During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time through the visual sensor mounted on the drone's gimbal.

[0049] S220. The first real-time images acquired in real time are sequentially input into the pre-trained target detection model to obtain the annotation results of the power distribution wires in each first real-time image output by the target detection model.

[0050] In a specific example, the object detection model can be understood as a dedicated algorithm built on the improved YOLOv11 framework, specifically optimized for power distribution line inspection scenarios. This model uses deep learning technology to accurately identify and locate targets such as conductors, tower heads, and tower bodies, outputting labeled results including target category and bounding box coordinates.

[0051] In this embodiment, the first real-time image captured by the UAV is input into a pre-trained dedicated target detection model for processing. This model is based on an improved YOLOv11 architecture and extracts image features through a deep convolutional neural network, enabling it to accurately identify power distribution wires in the image. The identified wires can be clearly marked with red bounding boxes.

[0052] S230. Based on the annotation results, calculate the percentage of the image width of the power distribution wires in each first real-time image.

[0053] In this embodiment, based on the wireline annotation results output by the object detection model, the image width proportion is determined by calculating the ratio of the wireline bounding box width to the image width. The specific processing procedure is as follows: first, the pixel width value w of the annotation box is obtained, and then divided by the total image width W to obtain the precise image width proportion value radio = w / W.

[0054] S240. If the current image width ratio of the power distribution wire is calculated to be greater than the first ratio threshold based on the current first real-time image, then control the drone to perform constant speed ascent flight.

[0055] The first proportional threshold can be understood as the upper limit of the width of the wire in the image (usually set to 20%).

[0056] In this embodiment, when the radio frequency of the current frame conductor exceeds a preset first proportional threshold, it indicates that the distance between the drone and the conductor is too close, posing a potential safety risk. At this time, the constant speed ascent control mechanism is immediately triggered, and the drone begins to ascend at a preset constant speed.

[0057] S250. If the current image width ratio of the power distribution wire is calculated to be less than the second proportional threshold based on the current first real-time image, then control the drone to perform constant speed descent flight, wherein the second proportional threshold is less than the first proportional threshold.

[0058] The second proportional threshold can be understood as representing the lower limit of the radio frequency of the conductor in the image (usually set to 15%).

[0059] In this embodiment, when the radio of the current frame conductor is detected to be lower than the preset second proportional threshold, it indicates that the distance between the UAV and the conductor is too far, affecting the inspection quality. At this time, the constant speed descent control mechanism is immediately triggered, and the UAV begins to descend at the preset constant speed.

[0060] S260. When the proportion of the image width of the power distribution conductor calculated for a preset number of consecutive first real-time images is between the first proportional threshold and the second proportional threshold, it is determined that the UAV has reached the target safe height matching the target tower.

[0061] In this embodiment, when the calculation results of the first real-time image for a consecutive preset number of frames (usually 10 frames) show that the radio of the conductor is maintained within the range of the first proportional threshold to the second proportional threshold, it is determined that the drone has reached a safe flight altitude matching the target tower and will no longer perform constant speed ascent or descent operations.

[0062] S270: After the UAV flies past the target tower at the target safe altitude, it continues to collect a second real-time image of the flight direction through the visual sensor.

[0063] S280. Based on the image width ratio of the power distribution wire in each second real-time image and the real-time positioning information of the UAV, control the UAV to take the target tower as the starting point and follow the height fluctuation of the power distribution wire in the flight direction to change the flight altitude.

[0064] The technical solution of this invention involves the following steps: During the altitude calibration phase before the UAV flies to the next target tower, a first real-time image along the flight path is acquired in real time by a visual sensor mounted on the gimbal. This image is then sequentially input into a pre-trained target detection model to obtain the annotation results of the power distribution wires in each frame and calculate the image width ratio. When the width ratio of the current frame is greater than a first ratio threshold, the UAV is controlled to perform a constant-speed ascent; when the width ratio is less than a second ratio threshold, a constant-speed descent is performed. This adjustment continues until the width ratio of multiple consecutive frames stabilizes within a preset range, confirming that the target safe altitude has been reached. After completing altitude calibration, the drone flies over the target tower at the safe altitude and continues to collect a second real-time image through visual sensors. Combining this with real-time positioning information, and taking the target tower as a starting point, the drone dynamically controls its flight altitude along the undulations of the conductor based on the changes in the width ratio and spatial relationship of the conductor in subsequent images. This achieves precise terrain-following flight. This novel AI vision-based terrain-following flight method for power distribution drones ensures that the drone can quickly and accurately adjust to the optimal inspection altitude through improved target detection algorithms and intelligent proportional threshold judgment mechanisms. At the same time, the deep fusion of visual perception and positioning data significantly reduces the need for manual intervention, enabling the drone to autonomously complete high-quality power inspection tasks under complex terrain conditions, greatly improving flight safety and operational efficiency.

[0065] Furthermore, based on the above embodiments, according to the image width ratio of the power distribution wire in each second real-time image and the real-time positioning information of the UAV, controlling the UAV to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution wire in the flight direction may further include:

[0066] If the width of the tower head of the next tower is detected to exceed the tower head ratio threshold in the current second real-time image, or if the image height of the bottom of the tower body of the next tower is detected to exceed the image height threshold, then the image height ratio of the top edge of the tower head or the top edge of the tower body is calculated.

[0067] If the height percentage of the image is less than or equal to the height percentage threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude.

[0068] If the image height percentage is greater than the height percentage threshold, then calculate the vertical offset angle of the image height percentage relative to the center point of the current second real-time image;

[0069] If the vertical offset angle is greater than the angle threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude.

[0070] While the drone is in tower ascent mode, the drone's horizontal speed is updated in real time based on the real-time calculated image height ratio until the drone is determined to have reached a safe altitude.

[0071] Generally, a height safety assessment mechanism will be activated when a drone detects during flight that the width of the top of the next tower exceeds a preset threshold (e.g., 15% of the image width) or the bottom of the tower exceeds an image height threshold (e.g., 25% of the bottom of the image). This process first calculates the height percentage of the top or bottom edge of the tower in the image. , where y is the ordinate of the top or upper edge of the tower, and H is the total height of the image. This parameter directly reflects the relative height relationship between the drone and the tower.

[0072] Generally, when a drone detects the next tower or tower top, it first determines the distance to the top edge of the recognition frame. If the elevation change is within 15%, the current height is considered unsafe, and the tower will enter ascent mode. If the percentage is greater than 15%, it is necessary to further calculate the vertical offset angle of the height percentage relative to the center point of the current second real-time image to determine whether it is necessary to enter the tower ascent mode. The vertical offset angle is... The calculation method is as follows:

[0073] First, define the image coordinate system, assuming the top Y-axis coordinate is 0 and the bottom coordinate is 1. Second, calculate... Position offset relative to the center of the screen By applying the azimuth function Math.atan2(), the vertical offset angle of the target point relative to the center of the screen can be calculated. Specifically, ,in It is a calculation point and the origin The angle (in radians) between the lines connecting them and the positive x-axis is a function of this angle. Because a safe altitude for the drone needs to be reserved, when... If the angle is greater than -4°, the current height is considered unsafe, and the tower will enter ascent mode.

[0074] Generally, entering the tower ascent mode will recalculate the drone's forward flight speed. The calculation process is as follows:

[0075] Set the drone's horizontal speed to The vertical velocity is ,when When less than 0.2, ,when When it is greater than 0.2, ,when When it is greater than 0.4, This indicates that the altitude has been raised to a safe level. Here, max(0, 0.4 - h_top) represents obtaining the maximum value between 0 and (0.4 - h_top).

[0076] Example 3

[0077] Figure 3 This is a flowchart of another AI vision-based method for ground-following flight of a power distribution drone, provided in Embodiment 3 of the present invention. This embodiment is based on and optimized from the above embodiments. Specifically, the operation of "controlling the drone to change its flight altitude by taking the target tower as the starting point and following the height fluctuations of the power distribution wire in the flight direction, based on the image width ratio of the power distribution wire in each second real-time image and the real-time positioning information of the drone" has been refined.

[0078] Correspondingly, such as Figure 3 As shown, the method includes:

[0079] S310: During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time by the visual sensor mounted on the drone gimbal.

[0080] S320. Based on the proportion of the power distribution wires in each first real-time image, dynamically adjust the drone's flight altitude until the drone reaches the target safe altitude that matches the target tower.

[0081] After the UAV flies past the target tower at the target safe altitude, the S330 continues to collect a second real-time image of the flight direction through the visual sensor.

[0082] S340. Based on the proportion of the width of the power distribution conductor in multiple consecutive second real-time images, determine the trend of the width change of the power distribution conductor starting from the target tower.

[0083] In this embodiment, after the UAV flies over the target tower, the width variation pattern of the conductor is analyzed by continuously acquiring multiple frames of second real-time images. A time-series analysis is performed on the conductor width percentage data based on a preset sliding window, calculating the linear regression slope of its changing trend. When the slope exceeds a positive threshold, it is determined to be an increasing trend; when it is below a negative threshold, it is determined to be a decreasing trend. This trend judgment process comprehensively considers the absolute value of the width percentage in the current frame and the changing direction of multiple consecutive frames, providing a reliable basis for subsequent flight mode decisions. The sliding window can be understood as a data processing technique in time series analysis; in this scheme, it specifically refers to a caching mechanism used to store and analyze conductor width percentage data in consecutive image frames.

[0084] S350. If the current image width ratio of the power distribution conductor is calculated to be greater than the third ratio threshold based on the current second real-time image, and the width change trend is an increasing trend, then the UAV is determined to enter the terrain-following flight ascent mode, and the UAV is controlled to ascend at a constant speed until the ascent control conditions are met.

[0085] The third proportional threshold can be understood as one of the trigger conditions for the drone's ascent mode. Together with the increasing width trend, it forms the judgment standard, ensuring the accuracy and safety of altitude adjustments. The ascent control termination condition can be understood as determining when to terminate the ascent operation and transition to a stable flight state, ensuring that the drone stops ascending in time when it reaches the optimal inspection altitude, avoiding excessive distance from the guide wire or entry into dangerous airspace. Specifically, this third proportional threshold can be preset according to actual conditions.

[0086] In this embodiment, when continuously analyzing multiple frames of the second real-time image, if the detected wire radio exceeds a preset third proportion threshold, and the width is confirmed to show a continuously increasing trend through sliding window calculation, the UAV is triggered to enter a specific flight adjustment mode. This judgment is based on rigorous time-series data analysis, requiring both the width proportion and the trend of change to be met simultaneously, ensuring the accuracy and reliability of flight altitude adjustment. After entering this mode, the UAV adjusts its altitude according to preset rules until subsequent detection data indicates that a safe flight state has been reached.

[0087] S360. If the current image width ratio of the power distribution conductor is less than the fourth ratio threshold and the width change trend is a decreasing trend, then the UAV is determined to enter the terrain-following flight descent mode, and the UAV is controlled to descend at a constant speed until the descent control conditions are met.

[0088] The fourth proportional threshold can be understood as one of the trigger conditions for the drone's descent mode, indicating that the drone has moved away from the power distribution line. The descent termination condition can be understood as determining when to stop the descent operation and enter a stable flight state, ensuring that the drone can stop descending in time when it reaches the optimal inspection altitude, avoiding excessive approach to the ground or power lines. Specifically, the third proportional threshold can be preset according to the actual situation.

[0089] In this embodiment, when continuously analyzing multiple frames of the second real-time image, if the radio wave of the guide wire is detected to be below a preset fourth proportional threshold, and the width is confirmed to show a continuously decreasing trend through sliding window calculation, the drone is triggered to enter a specific descent adjustment mode. After entering this mode, the drone adjusts its altitude according to preset rules until subsequent detection data indicates that a safe flight state has been reached.

[0090] S370. Based on the real-time positioning information of the UAV, obtain the target vertical flight speed when the UAV meets the conditions for ending ascent control or ending descent control.

[0091] In this embodiment, when the UAV completes its ascent or descent mode and meets the termination control conditions, the target's vertical flight speed is calculated using real-time acquired positioning information. This process integrates elevation data provided by GNSS (Global Navigation Satellite System) and the relative position of the guide wire measured by visual sensors, and determines the optimal vertical speed parameters based on the spatial relationship between the current flight state and the target's safe altitude.

[0092] S380: While controlling the UAV to fly at the target vertical speed, it continues to acquire third real-time images in the flight direction through visual sensors.

[0093] The third real-time image can be understood as the flight direction image data continuously collected by the visual sensor after the UAV completes altitude calibration and enters the terrain-following flight phase.

[0094] S390. When the current image width ratio of the power distribution conductor is calculated based on the current third real-time image and meets the conditions for fine-tuning the flight altitude, calculate the real-time control parameters of the UAV.

[0095] The altitude fine-tuning condition can be understood as the criterion for triggering a small altitude adjustment during the cruise phase of the drone to maintain the optimal relative position with the guide wire. This condition is based on visual analysis results; fine-tuning control is initiated when the guide wire radio is detected to deviate from the safe range but has not reached the mode switching threshold.

[0096] Real-time control parameters can be understood as a set of adjustment commands dynamically generated by the UAV during terrain-following flight, used to precisely control flight altitude and speed.

[0097] In this embodiment, when the UAV is in terrain-following flight mode, fine-tuning of the flight altitude is triggered by analyzing the radio changes of the guide wires in the third real-time image. This process is based on a preset fine-tuning condition threshold. When a radio deviation from the safe range is detected but has not reached the mode switching threshold, a small-scale altitude correction is initiated. The fine-tuning control adopts a progressive strategy, dynamically calculating the required control parameters based on the degree of difference between the current deviation and the target value, ensuring a smooth transition of the flight trajectory to the ideal state.

[0098] S3100: Adjust the control input of the PID controller in the UAV according to the real-time control parameters of the UAV.

[0099] Among them, the PID controller can be understood as one of the control algorithms in the field of control. It achieves precise regulation of the controlled object by organically combining the three control actions of proportional (P), integral (I), and derivative (D).

[0100] In this embodiment, the parameter adjustment process of the PID controller is a dynamic optimization based on real-time calculated flight state deviations. When the UAV determines that altitude fine-tuning is required based on the third real-time image analysis results, it dynamically calculates and updates the three gain parameters (proportional coefficient Kp, integral coefficient Ki, and derivative coefficient Kd) of the PID controller according to the deviation between the current wire width ratio and the target value, the deviation change trend, and the historical cumulative error.

[0101] The technical solution of this invention involves the following steps: During the altitude calibration phase before the UAV flies to the next target tower, real-time images are acquired using a gimbal vision sensor. These images are then processed by a target detection model to obtain the conductor width ratio. When the ratio exceeds a first threshold, a constant-speed ascent is triggered; when it falls below a second threshold, a constant-speed descent is triggered, until multiple consecutive frames of data stabilize within a safe range, confirming the altitude calibration is complete. After altitude calibration, the UAV flies over the target tower at this safe altitude and continues to acquire second real-time images along the flight direction using the vision sensor. Based on the conductor width ratio in multiple consecutive second real-time images, the trend of conductor width variation starting from the target tower is determined. If the conductor width ratio in the current second real-time image is greater than a third proportional threshold and shows an increasing trend, the UAV enters a terrain-following ascent mode and controls a constant-speed ascent; if the width ratio is less than a fourth proportional threshold and shows a decreasing trend, the UAV enters a terrain-following descent mode and controls a constant-speed descent. The target's vertical flight speed at which the control termination conditions are met is obtained based on real-time positioning information, and third real-time images are continuously acquired during flight at this speed. When the proportion of conductor width in the third real-time image meets the fine-tuning conditions, the real-time control parameters are calculated and the control quantity of the PID controller is adjusted. This new AI vision-based distribution network drone terrain-following flight method achieves high-precision altitude control through trend analysis of continuous multi-frame images and dynamic adjustment of the PID controller, while greatly reducing hardware costs and providing reliable technical support for intelligent inspection of distribution networks.

[0102] Optionally, based on the above embodiments, determining the width variation trend of the power distribution conductor starting from the target tower, according to the image width proportion of the power distribution conductor in multiple consecutive second real-time images, may include:

[0103] The image width ratio of the power distribution conductor in a series of consecutive second real-time images of a preset number n is input into a pre-constructed Kalman filter for filtering to obtain multiple standard image width ratios;

[0104] Store the width proportions of each standard image sequentially into the window data cache queue;

[0105] Based on the width proportions of each standard image stored in the window data cache queue, the linear regression slope is calculated using the following formula. ;

[0106]

[0107] Where sum represents the accumulation operation. This represents the sum of the sequence numbers of all frames within the window. This represents the sum of the proportions of the wire width across all frames of the image. This represents the sum of the squares of all frame numbers. This represents the sum of the products of all frame numbers and their corresponding width proportions, where i is the index of the width proportion of each standard image in the window data buffer queue. The percentage of the standard image width corresponding to the i-th index in the window data cache queue. For covariance components, For variance components;

[0108] When the slope of the linear regression is greater than the first slope threshold, the trend of width change is determined to be an increasing trend.

[0109] When the slope of the linear regression is less than the second slope threshold, the trend of width change is determined to be a decreasing trend.

[0110] Generally, during the drone's terrain-following flight, a preset number of n consecutive second real-time images (usually 30 frames) are fed into a pre-built Kalman filter for noise reduction. First, before each frame of tracking wire data arrives, the tracking wire width w is input, processed by the Kalman filter, and then normalized using the following formula:

[0111]

[0112] ;

[0113] in, This indicates an estimate of the covariance. This is expressed as the process noise covariance. This is expressed as the optimal estimate of the conductor width. This is expressed as the measurement noise covariance. It is expressed as the sum of the process noise covariance and the estimated covariance.

[0114] At the same time, update the estimated covariance. Adaptive updates Based on this The wire width can be updated accordingly. The width of the processed wire Place it into a window listW of size 30 for subsequent calculations.

[0115] Using the data in the listW window, a trend calculation is performed on the data within the window using linear regression. The slope of the linear regression is calculated based on the following formula:

[0116]

[0117] Where sum represents the accumulation operation. This represents the sum of the sequence numbers of all frames within the window. This represents the sum of the proportions of the wire width across all frames of the image. This represents the sum of the squares of all frame numbers. This represents the sum of the products of all frame numbers and their corresponding width proportions, where i is the index of the width proportion of each standard image in the window data buffer queue. The percentage of the standard image width corresponding to the i-th index in the window data cache queue. For covariance components, For variance components;

[0118] Generally, in UAV terrain-following flight control, the changing trend of the proportion of wire width in consecutive frames of images is analyzed by linear regression. When the calculated regression slope exceeds the first slope threshold (usually set to +2), it is determined that the width of the wire in the image is continuously increasing, indicating that the distance between the UAV and the wire is decreasing, and the width change shows an increasing trend. Conversely, when the regression slope is lower than the second slope threshold (usually set to -2), it is determined that the wire width is continuously decreasing, the distance between the UAV and the wire is increasing, and the width change shows a decreasing trend.

[0119] Controlling the drone to ascend at a constant speed until the conditions for ending ascent are met includes:

[0120] During the process of controlling the drone to ascend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the third ratio threshold calculated based on the current second real-time image meets the first difference value condition, then it is determined that the condition for ending the ascent control is met.

[0121] Controlling the drone to descend at a constant speed until the descent termination control conditions are met includes:

[0122] During the process of controlling the drone to descend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the fourth ratio threshold calculated for the current second real-time image meets the second difference value condition, then it is determined that the descent control condition is met.

[0123] Accordingly, based on the UAV's real-time positioning information, the target vertical flight speed of the UAV when the conditions for ending ascent control or descent control are met is obtained, including:

[0124] When the UAV meets the conditions for ending ascent control or ending descent control, acquire the first real-time positioning information of the UAV at the first position and the second real-time positioning information at the second position, collected at two consecutive time points.

[0125] Based on the latitude and longitude differences between the first and second positioning information, as well as the Earth's radius, the ground distance between the first and second positions is calculated.

[0126] Based on the altitude difference between the first and second positioning information and the ground distance, the slope value between the first and second positions is calculated, and the target vertical flight speed of the UAV is calculated based on the calculated slope value.

[0127] The first difference value condition can be understood as the quantitative judgment standard for the drone to end its ascent control. It refers to the difference between the current width ratio of the conductor (radio) and the third ratio threshold, which must reach a preset range and continuously meet the conditions for multi-frame verification. Similarly, the second difference value condition is the quantitative judgment standard for the drone to end its descent control. These first and second difference value conditions can be selected based on the actual situation.

[0128] Generally, during a constant-speed ascent or descent, the drone determines whether to terminate the current control mode by analyzing the difference between the proportion of the wire width in the second real-time image and a preset threshold. This mechanism employs a dual-verification strategy, considering both absolute deviation and continuous stability.

[0129] Generally, when a drone meets the ascent or descent termination conditions, it records GNSS positioning data for two locations, A1 and A2, collected at two consecutive time points (t1 and t2): A1 = (lat1, lon1, alt1) and A2 = (lat2, lon2, alt2). Here, lat1, lon1, and alt1 represent the latitude, longitude, and altitude information of location A1 collected at time t1; lat2, lon2, and alt2 represent the latitude, longitude, and altitude information of location A2 collected at time t2.

[0130] The ground distance is calculated using the Haversine formula, as shown below:

[0131]

[0132] in, This is the latitude information of location point A1, lat1, converted to radians.

[0133] This involves converting the dimension lat2 of location point A2 into latitude information in radians. It is the difference in latitude between points A1 and A2 in radians. R is the difference in longitude between points A1 and A2, where lon1 and lon2 are the longitudes of the points A1 and A2 respectively. R is the Earth's radius, d is the ground distance between points A1 and A2, and a and c are intermediate variables. a represents the semi-sine function value of the angular distance between the two points, and c represents the central angle (in radians) between the two points.

[0134] Then, by calculating the altitude difference h = alt2 - alt1, the slope k between points A1 and A2 is obtained using the arctangent function atan().

[0135]

[0136] At this point, the known horizontal speed of the drone is Vertical velocity At the same time, the GNSS data at this moment is recorded and used as the initial GNSS data for the next slope k update calculation, with each update interval being at least 3 seconds.

[0137] Optionally, based on the above embodiments, when the current image width ratio of the power distribution conductor is calculated to meet the flight altitude fine-tuning conditions according to the current third real-time image, the calculation of the real-time control parameters of the UAV may include:

[0138] Calculate the difference between the current image width percentage of the power distribution conductor calculated based on the current third real-time image and the previous image width percentage of the power distribution conductor calculated based on the previous third real-time image.

[0139] If the percentage difference exceeds a preset difference threshold, then based on the percentage difference, the proportional control quantity, integral control quantity, and derivative control quantity corresponding to the PID controller in the UAV are calculated as real-time control parameters.

[0140] Generally, flight status deviation is assessed by comparing the change in the proportion of the conductor width in two consecutive third real-time images. The formula for calculating the proportion difference Δradio is:

[0141]

[0142] in, This indicates the width of the wire in pixels in the current third real-time image. This indicates the width of the wire in pixels of the previous third real-time image.

[0143] Generally, when the difference in proportion exceeds a preset difference threshold, a PID controller will be used to compensate for the vertical speed. The specific implementation of the PID algorithm is as follows:

[0144] The proportional control quantity is obtained by multiplying the current value of the deviation by the proportional coefficient Kp. The proportional element's function is to provide a rapid response to the deviation; the larger the deviation, the larger the control quantity. The proportional control quantity is derived from the following formula. :

[0145] .

[0146] The integral control quantity is obtained by integrating the deviation, based on the following formula. :

[0147] Where n represents the time series and N represents the current time. Indicates the integration time interval. This represents the deviation value at time n. This represents the integral coefficient. The function of the integral element is to eliminate steady-state error. Even when the deviation is small, through continuous accumulation, the control variable can reach a sufficient value to eliminate the error.

[0148] The rate of change of the deviation is calculated, and the differential control quantity is obtained based on the following formula. Calculation method:

[0149]

[0150] in, The differential coefficients are... This represents the deviation value at time n. This represents the deviation value at the previous time step (n-1). The function of the differential element is to predict the trend of deviation changes, adjust the control quantity in advance, increase the stability of the system, and avoid overshoot.

[0151] Adding the proportional, integral, and derivative control values ​​together yields the final control value:

[0152]

[0153] Ultimately, the drone's vertical speed is This enables a terrain-following flight method based on the fusion of traverse lines and GNSS data.

[0154] Example 4

[0155] Figure 4 This is a schematic diagram of the structure of a data distribution network drone based on AI vision, provided in Embodiment 4 of the present invention. Figure 4 As shown, the device includes:

[0156] The altitude calibration module 410 is used to acquire the first real-time image in the flight direction in real time through the visual sensor mounted on the drone gimbal during the altitude calibration stage before the drone flies to the next target tower.

[0157] The dynamic height adjustment module 420 is used to dynamically adjust the flight height of the UAV based on the image width ratio of the power distribution wire in each first real-time image until the UAV reaches the target safe height that matches the target tower.

[0158] The target tower transition module 430 is used to continue to acquire a second real-time image in the flight direction through a visual sensor after the UAV has flown past the target tower at the target safe altitude.

[0159] The follow-flight module 440 is used to control the UAV to change its flight altitude by following the height fluctuations of the power distribution wire in the flight direction, based on the image width ratio of the power distribution wire in each second real-time image and the real-time positioning information of the UAV, starting from the target tower.

[0160] The technical solution of this invention involves a height calibration phase before the drone flies to the next target tower. A first real-time image along the flight path is acquired in real-time using a visual sensor mounted on the drone's gimbal. The drone's flight altitude is dynamically adjusted based on the width ratio of the power distribution wires in each of the first real-time images until a target safe altitude matching the target tower is reached. After the drone flies past the target tower at the target safe altitude, a second real-time image along the flight path is acquired in real-time using the visual sensor. Based on the width ratio of the power distribution wires in each of the second real-time images and the drone's real-time positioning information, the drone is controlled to change its flight altitude from the target tower, following the height fluctuations of the power distribution wires along the flight path, thus achieving autonomous terrain-following flight. This novel AI vision-based terrain-following flight method for power distribution drones acquires wire images in real-time using a visual sensor and analyzes the width ratio, enabling the drone to accurately follow the spatial direction of the wires without relying on lidar or pre-installed terrain data, significantly improving the safety and economy of power grid inspection.

[0161] Based on the above embodiments, the dynamic height adjustment module 420 is specifically used for:

[0162] The first real-time images acquired in real time are sequentially input into the pre-trained target detection model to obtain the annotation results of the power distribution wires in each first real-time image output by the target detection model;

[0163] Based on the annotation results, the image width ratio of the power distribution wires in each first real-time image is calculated;

[0164] If the current image width ratio of the power distribution wire is calculated to be greater than the first ratio threshold based on the current real-time image, then the drone is controlled to ascend at a constant speed.

[0165] If the current image width ratio of the power distribution wire is calculated to be less than the second ratio threshold based on the current first real-time image, then the drone is controlled to descend at a constant speed, wherein the second ratio threshold is less than the first ratio threshold.

[0166] When the proportion of the image width of the power distribution conductor calculated for a preset number of consecutive first real-time images is between the first proportional threshold and the second proportional threshold, it is determined that the UAV has reached the target safe height matching the target tower.

[0167] Based on the above embodiments, the follow-flight module 440 may include:

[0168] The trend analysis submodule is used to determine the width change trend of the power distribution conductor starting from the target tower based on the proportion of the image width of the power distribution conductor in multiple consecutive second real-time images.

[0169] The ascent trigger submodule is used to determine that the UAV enters the terrain-following ascent mode if the current image width ratio of the power distribution conductor is calculated to be greater than the third ratio threshold based on the current second real-time image, and the width change trend is an increasing trend. Then, the UAV is controlled to ascend at a constant speed until the ascent control conditions are met.

[0170] The descent trigger submodule is used to determine that the UAV enters the terrain-following descent mode if the current image width ratio of the power distribution conductor is less than the fourth ratio threshold and the width change trend is a decreasing trend, based on the current second real-time image. It then controls the UAV to descend at a constant speed until the descent control conditions are met.

[0171] The vertical speed submodule is used to obtain the target vertical flight speed of the UAV when the conditions for ending ascent control or ending descent control are met, based on the UAV's real-time positioning information.

[0172] The terrain-following detection submodule is used to continue to acquire a third real-time image along the flight direction through a visual sensor while controlling the UAV to fly at the target vertical speed.

[0173] The fine-tuning submodule is used to calculate the real-time control parameters of the UAV when the current image width ratio of the power distribution wire meets the flight altitude fine-tuning conditions based on the current third real-time image.

[0174] The controller update submodule is used to adjust the control input of the PID controller in the UAV according to the UAV's real-time control parameters.

[0175] Based on the above embodiments, the trend analysis submodule is specifically used for:

[0176] The image width ratio of the power distribution conductor in a series of consecutive second real-time images of a preset number n is input into a pre-constructed Kalman filter for filtering to obtain multiple standard image width ratios;

[0177] Store the width proportions of each standard image sequentially into the window data cache queue;

[0178] Based on the width proportions of each standard image stored in the window data cache queue, the linear regression slope is calculated using the following formula. ;

[0179]

[0180] Where sum represents the accumulation operation. This represents the sum of the sequence numbers of all frames within the window. This represents the sum of the proportions of the wire width across all frames of the image. This represents the sum of the squares of all frame numbers. This represents the sum of the products of all frame numbers and their corresponding width proportions, where i is the index of the width proportion of each standard image in the window data buffer queue. The percentage of the standard image width corresponding to the i-th index in the window data cache queue. For covariance components, For variance components;

[0181] When the slope of the linear regression is greater than the first slope threshold, the trend of width change is determined to be an increasing trend.

[0182] When the slope of the linear regression is less than the second slope threshold, the trend of width change is determined to be a decreasing trend.

[0183] Based on the above embodiments, the rise trigger submodule is specifically used for:

[0184] During the process of controlling the drone to ascend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the third ratio threshold calculated based on the current second real-time image meets the first difference value condition, then it is determined that the condition for ending the ascent control is met.

[0185] Based on the above embodiments, the descent trigger submodule is specifically used for:

[0186] During the process of controlling the drone to descend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the fourth ratio threshold calculated for the current second real-time image meets the second difference value condition, then it is determined that the descent control condition is met.

[0187] Based on the above embodiments, the vertical velocity submodule is specifically used for:

[0188] When the UAV meets the conditions for ending ascent control or ending descent control, acquire the first real-time positioning information of the UAV at the first position and the second real-time positioning information at the second position, collected at two consecutive time points.

[0189] Based on the latitude and longitude differences between the first and second positioning information, as well as the Earth's radius, the ground distance between the first and second positions is calculated.

[0190] Based on the altitude difference between the first and second positioning information and the ground distance, the slope value between the first and second positions is calculated, and the target vertical flight speed of the UAV is calculated based on the calculated slope value.

[0191] Based on the above embodiments, the fine-tuning submodule is specifically used for:

[0192] Calculate the difference between the current image width percentage of the power distribution conductor calculated based on the current third real-time image and the previous image width percentage of the power distribution conductor calculated based on the previous third real-time image.

[0193] If the percentage difference exceeds a preset difference threshold, then based on the percentage difference, the proportional control quantity, integral control quantity, and derivative control quantity corresponding to the PID controller in the UAV are calculated as real-time control parameters.

[0194] Furthermore, based on the above embodiments, the follow-flight module 440 may also include: [further details to be added].

[0195] If the width of the tower head of the next tower is detected to exceed the tower head ratio threshold in the current second real-time image, or if the image height of the bottom of the tower body of the next tower is detected to exceed the image height threshold, then the image height ratio of the top edge of the tower head or the top edge of the tower body is calculated.

[0196] If the height percentage of the image is less than or equal to the height percentage threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude.

[0197] If the image height percentage is greater than the height percentage threshold, then calculate the vertical offset angle of the image height percentage relative to the center point of the current second real-time image;

[0198] If the vertical offset angle is greater than the angle threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude.

[0199] While the drone is in tower ascent mode, the drone's horizontal speed is updated in real time based on the real-time calculated image height ratio until the drone is determined to have reached a safe altitude.

[0200] The AI ​​vision-based distribution network drone terrain-following flight device provided in this embodiment of the invention can execute the AI ​​vision-based distribution network drone terrain-following flight method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0201] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0202] Example 5

[0203] Figure 5 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0204] like Figure 5 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0205] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0206] Processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as performing an AI vision-based distribution network drone terrain-following flight method as described in any embodiment of the present invention, namely:

[0207] During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time by the visual sensor mounted on the drone gimbal.

[0208] Based on the proportion of the image width of the power distribution wires in each first real-time image, the flight altitude of the UAV is dynamically adjusted until the UAV reaches the target safe altitude that matches the target tower.

[0209] After the drone flies past the target tower at the target safe altitude, it continues to collect a second real-time image of the flight direction through the visual sensor.

[0210] Based on the proportion of the power distribution line's width in each second real-time image and the UAV's real-time positioning information, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution line in the flight direction.

[0211] In some embodiments, a method for AI vision-based distribution network drone terrain-following flight as described in any of the embodiments of the present invention can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for AI vision-based distribution network drone terrain-following flight as described above in any of the embodiments of the present invention can be performed. Alternatively, in other embodiments, processor 11 can be configured by any other suitable means (e.g., by means of firmware) to perform the method for AI vision-based distribution network drone terrain-following flight as described in any of the embodiments of the present invention.

[0212] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0213] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0214] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0215] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0216] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0217] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0218] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0219] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for ground-following flight of a power distribution drone based on AI vision, characterized in that, The method includes: During the altitude calibration phase before the drone flies to the next target tower, the first real-time image in the flight direction is acquired in real time by the visual sensor mounted on the drone gimbal. Based on the proportion of the image width of the power distribution wires in each first real-time image, the flight altitude of the UAV is dynamically adjusted until the UAV reaches the target safe altitude that matches the target tower. After the drone flies past the target tower at the target safe altitude, it continues to collect a second real-time image of the flight direction through the visual sensor. Based on the proportion of the power distribution wire in each second real-time image and the real-time positioning information of the UAV, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution wire in the flight direction. Specifically, based on the image width ratio of the power distribution wires in each second real-time image and the real-time positioning information of the UAV, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution wires in the flight direction, including: Based on the image width proportion of the power distribution conductor in multiple consecutive second real-time images, the width change trend of the power distribution conductor starting from the target tower is determined. If the current image width proportion of the power distribution conductor calculated for the current second real-time image is greater than the third proportion threshold, and the width change trend is an increasing trend, then the UAV is determined to enter the terrain-following flight ascent mode, and the UAV is controlled to ascend at a constant speed until the ascent control condition is met. If the current image width proportion of the power distribution conductor calculated for the current second real-time image is less than the fourth proportion threshold, and the width change trend is a decreasing trend, then the UAV is determined to enter the terrain-following flight descent mode, and the UAV is controlled to descend at a constant speed until the descent control condition is met. Specifically, based on the proportion of the power distribution conductor's width in multiple consecutive second real-time images, the trend of the power distribution conductor's width variation starting from the target tower is determined, including: The image width proportions of power distribution conductors in a predetermined number n consecutive second real-time images are input into a pre-constructed Kalman filter for filtering to obtain multiple standard image width proportions. Each standard image width proportion is then stored sequentially in a window data cache queue. Based on the standard image width proportions stored in the window data cache queue, the linear regression slope is calculated using the following formula. ; ； Where sum represents the accumulation operation. This represents the sum of the sequence numbers of all frames within the window. This represents the sum of the proportions of the wire width across all frames of the image. This represents the sum of the squares of all frame numbers. This represents the sum of the products of all frame numbers and their corresponding width proportions, where i is the index of the width proportion of each standard image in the window data buffer queue. The percentage of the standard image width corresponding to the i-th index in the window data cache queue. For covariance components, The variance component is defined as follows: when the slope of the linear regression is greater than the first slope threshold, the trend of width change is determined to be an increasing trend; when the slope of the linear regression is less than the second slope threshold, the trend of width change is determined to be a decreasing trend.

2. The method according to claim 1, characterized in that, Based on the image width ratio of the power distribution wires in each of the first real-time images, the drone's flight altitude is dynamically adjusted until the drone reaches the target safe altitude matching the target tower, including: The first real-time images acquired in real time are sequentially input into the pre-trained target detection model to obtain the annotation results of the power distribution wires in each first real-time image output by the target detection model; Based on the annotation results, the image width ratio of the power distribution wires in each first real-time image is calculated; If the current image width ratio of the power distribution wire is calculated to be greater than the first ratio threshold based on the current real-time image, then the drone is controlled to ascend at a constant speed. If the current image width ratio of the power distribution wire is calculated to be less than the second ratio threshold based on the current first real-time image, then the drone is controlled to descend at a constant speed, wherein the second ratio threshold is less than the first ratio threshold. When the proportion of the image width of the power distribution conductor calculated for a preset number of consecutive first real-time images is between the first proportional threshold and the second proportional threshold, it is determined that the UAV has reached the target safe height matching the target tower.

3. The method according to claim 1, characterized in that, Based on the image width proportion of the power distribution wires in each second real-time image and the real-time positioning information of the UAV, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution wires in the flight direction, including: Based on the real-time positioning information of the UAV, obtain the target vertical flight speed when the conditions for ending ascent control or descent control are met. While controlling the drone to fly at the target vertical speed, the third real-time image in the flight direction is continuously acquired through the visual sensor; When the current image width ratio of the power distribution conductor meets the flight altitude fine-tuning conditions based on the current third real-time image, calculate the real-time control parameters of the UAV. The control input of the PID controller in the UAV is adjusted according to the real-time control parameters of the UAV.

4. The method according to claim 1, characterized in that, Controlling the drone to ascend at a constant speed until the conditions for ending ascent are met includes: During the process of controlling the drone to ascend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the third ratio threshold calculated based on the current second real-time image meets the first difference value condition, then it is determined that the condition for ending the ascent control is met. Controlling the drone to descend at a constant speed until the descent termination control conditions are met includes: During the process of controlling the drone to descend at a constant speed, if the difference between the current image width ratio of the power distribution wire and the fourth ratio threshold calculated for the current second real-time image meets the second difference value condition, then it is determined that the descent control condition is met. Accordingly, based on the UAV's real-time positioning information, the target vertical flight speed of the UAV when the conditions for ending ascent control or descent control are met is obtained, including: When the UAV meets the conditions for ending ascent control or ending descent control, acquire the first real-time positioning information of the UAV at the first position and the second real-time positioning information at the second position, collected at two consecutive time points. Based on the latitude and longitude differences between the first and second positioning information, as well as the Earth's radius, the ground distance between the first and second positions is calculated. Based on the altitude difference between the first and second positioning information and the ground distance, the slope value between the first and second positions is calculated, and the target vertical flight speed of the UAV is calculated based on the calculated slope value.

5. The method according to claim 3, characterized in that, When the current image width ratio of the power distribution conductor, calculated based on the current third real-time image, meets the conditions for fine-tuning the flight altitude, the real-time control parameters of the UAV are calculated, including: Calculate the difference between the current image width percentage of the power distribution conductor calculated based on the current third real-time image and the previous image width percentage of the power distribution conductor calculated based on the previous third real-time image. If the percentage difference exceeds a preset difference threshold, then based on the percentage difference, the proportional control quantity, integral control quantity, and derivative control quantity corresponding to the PID controller in the UAV are calculated as real-time control parameters.

6. The method according to any one of claims 1-5, characterized in that, Based on the image width ratio of the power distribution wires in each second real-time image and the real-time positioning information of the UAV, the UAV is controlled to change its flight altitude by starting from the target tower and following the height fluctuations of the power distribution wires in the flight direction. This also includes: If the width of the tower head of the next tower exceeds the tower head proportion threshold in the current second real-time image, or if the image height of the bottom of the tower body of the next tower exceeds the image height threshold, then the image height proportion of the top edge of the tower head or the top edge of the tower body is calculated. ; If the height percentage of the image is less than or equal to the height percentage threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude. If the image height percentage is greater than a height percentage threshold, then calculate the vertical offset angle of the image height percentage relative to the center point of the current second real-time image. ; ,in, , It is a calculation The offset of the position relative to the center of the screen. It is a calculation point and the origin A function of the angle in radians between the lines connecting them relative to the positive x-axis; If the vertical offset angle is greater than the angle threshold, the drone is controlled to enter the tower ascent mode to increase the drone's flight altitude. While the drone is in tower ascent mode, the drone's horizontal speed is updated in real time based on the real-time calculated image height ratio until the drone is determined to have reached a safe altitude.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the AI ​​vision-based distribution network drone terrain-following flight method according to any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the AI ​​vision-based method for the ground-following flight of a power distribution drone as described in any one of claims 1-6.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the AI ​​vision-based method for the ground-following flight of a power distribution drone according to any one of claims 1-6.

Images