A method for multi-transmitter unmanned aerial vehicle laser remote charging

Abstract

The application discloses a kind of multi-transmitting end unmanned plane laser remote charging method, the method includes: in transmitting end, 3 lasers are used parallel operation, establish multi-transmitting end unmanned plane and laser remote charging system of tracking and sighting;Real-time cooling is carried out to laser using air-cooled heat sink device simultaneously;Multi-transmitting end unmanned plane and laser remote charging system establishes a set of overall tracking and sighting device, is installed in one of turntable, tracking and sighting device returns the position information of laser to host computer by the way of binocular vision positioning, calculates corresponding offset data transmission to two driven turntables, and driven turntable is rotated according to instruction, the common tracking of three lasers is realized;Solve the technical problems, such as the energy loss caused by losing target due to laser spot and receiving end photovoltaic array area is smaller, unmanned plane flight speed is faster, the tracking accuracy and real-time response speed of transmitting end turntable have higher requirement, otherwise it will appear lose target, leading to energy loss, and unmanned plane cannot complete long time endurance.

Description

Technical Field

[0001] This invention belongs to the field of drone charging, and particularly relates to a method for remote laser charging of multi-transmitter drones. Background Technology

[0002] my country's power grid is distributed across the country, with some power lines located in mountainous areas, making manual maintenance too costly. With the development and widespread adoption of drone photography technology, automated drone inspections are increasingly used in the maintenance of power transmission lines, leading to a growing demand for efficient drone inspections. However, current drone autopilot systems require significant human intervention. This is due not only to the immaturity of autopilot technology but also to the critical limitation of drone battery life, which restricts automated inspection efficiency. Current lithium-ion battery-powered power transmission inspection drones typically have an effective operating time of no more than 30 minutes, hindering long-distance automated inspections and preventing a rapid improvement in efficiency. In the complex working environment of high-voltage electrical equipment, conventional methods for replenishing batteries are unsuitable, making drone power supply a major obstacle to continuous inspections.

[0003] Laser, as a medium for long-distance energy transmission, has advantages such as good directionality and long transmission distance. Using laser to provide real-time power to drones operating at long distances, and applying this technology to drone inspections within power grids, will make a significant contribution to monitoring power grid facilities and preventing catastrophic accidents.

[0004] However, traditional laser wireless power transmission technology is often limited by low transmission efficiency and severe heat dissipation problems of the laser, resulting in limited laser transmission power. Furthermore, laser transmitters typically require bulky and heavy heat dissipation devices to assist in their normal operation. In addition, due to the small area of ​​the laser spot and the photovoltaic array at the receiver, and the high flight speed of drones, high requirements are placed on the tracking accuracy and real-time response speed of the transmitter turntable; otherwise, target loss will occur, leading to energy loss and preventing the drone from achieving long-term endurance. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for remote laser charging of multi-transmitter UAVs, so as to solve the technical problems that, due to the small area of ​​the laser spot and the photovoltaic array of the receiving end, the high flight speed of the UAV, the tracking accuracy and real-time response speed of the transmitting end turntable are required, otherwise the target will be lost, resulting in energy loss and the UAV will not be able to complete long-term flight.

[0006] The technical solution of this invention is:

[0007] A method for remote laser charging of a multi-transmitter UAV includes: using three lasers operating in parallel at the transmitting end to establish a multi-transmitter UAV tracking and aiming laser remote charging system; simultaneously using an air-cooling device to dissipate heat from the lasers in real time; establishing an integrated tracking and aiming device within the multi-transmitter UAV tracking and aiming laser remote charging system, which is installed on one of the turntables; the tracking and aiming device returns the position information of the lasers to the host computer through binocular vision positioning, calculates the corresponding offset data, and transmits it to two slave turntables; the slave turntables rotate according to instructions to achieve joint tracking of the three laser beams.

[0008] The multi-transmitter UAV tracking and aiming laser remote charging system includes: a laser system, a tracking and aiming system, and a photovoltaic receiving system.

[0009] The tracking and aiming system includes two cameras, an image acquisition card, a computer, control circuitry, and servo motors. It is used to monitor the drone's flight status in real time and provide position information. Finally, based on the position information, it controls the rotation of the servo motors to achieve tracking and aiming.

[0010] The laser system consists of three lasers, which are synchronously tracked by the turntable's main control system, enabling all three lasers to be powered simultaneously.

[0011] The photovoltaic receiving system includes a drone and photovoltaic panels. The photovoltaic panels are mounted on the drone's landing gear via brackets and can be freely adjusted at a 30° angle to facilitate the photovoltaic array receiving laser light.

[0012] The tracking and aiming device includes two cameras mounted on either side of an optical collimating lens to capture images of the drone in flight from different angles. The captured images are then transmitted to an image acquisition card, which transmits the real-time image information to a computer. The computer then calculates the drone's position information based on two visual positioning theories.

[0013] The tracking and aiming device uses a binocular vision positioning method including:

[0014] Step 1: Real-time scene image acquisition: Binocular images are acquired through two cameras to form real-time images of the scene;

[0015] Step 2, Binocular camera calibration: Obtain the intrinsic and extrinsic parameters of the two cameras respectively, generate a mapping model from image coordinates to world coordinates, and form a calibration model for the binocular camera;

[0016] Step 3, Image Preprocessing: Using the acquired real-time scene images and the binocular camera calibration model, image preprocessing is performed on the acquired real-time scene images. Image preprocessing includes neighborhood averaging preprocessing and filtering preprocessing.

[0017] Step 4, Image Feature Point Matching: The ORB algorithm is used to detect features in the images from the left and right cameras. The detected feature points are matched according to the matching algorithm, and the matched feature point pairs are recorded.

[0018] Step 5: Calculate the disparity based on the matched feature point pairs, and obtain the three-dimensional coordinates of the matched point pairs in the camera coordinate system according to the principle of binocular vision ranging.

[0019] Step 5: Calculate the camera pose using the PnP or ICP algorithm based on the 3D coordinates and image coordinates of the matching points;

[0020] Step 6: Determine the current position and attitude of the drone based on the relative positional relationship between the camera and the drone;

[0021] Step 7: Correct the positioning error by drawing on the ideas of visual SLAM and beacon recognition and positioning;

[0022] Step 8: The UAV continuously moves and acquires position information in real time to obtain the global pose information of the UAV.

[0023] The method of returning the laser's position information to the host, calculating the corresponding offset data, and transmitting it to the two slave turntables, which then rotate according to the instructions, to achieve simultaneous tracking of the three laser beams includes: the computer transmitting the acquired position information to the PLC control circuit, controlling the servo motors to rotate and achieve tracking; since the position information of the UAV relative to the middle laser is known, and the horizontal distance between the two side lasers and the middle laser is set according to specific requirements, the offset angle between the two side lasers and the laser on the middle main control turntable is calculated according to the cosine theorem, and the offset angle instruction is transmitted to the PLC control circuit of the two side lasers to control the servo motors to rotate, thereby achieving synchronous tracking of the three lasers.

[0024] In actual operation, if more than one drone is detected in the camera's visual interface, the real-time battery information of the drone is transmitted to the transmitter using the wireless communication device between the drone and the transmitter. Then, the transmitter analyzes and calculates the battery level of each drone to obtain the optimal charging solution, prioritizing charging the drone with low battery.

[0025] The optimal charging solution is generated as follows:

[0026] (1) By combining binocular visual positioning with monocular camera image recognition positioning technology, the position and energy status information of the UAV are obtained and stored in the computer;

[0027] (2) When information about one UAV is obtained, the rotation of three lasers is controlled by the obtained UAV location information, so that the three lasers charge the UAV at the same time.

[0028] (3) When information of two drones is obtained, the three lasers are controlled to rotate by the obtained drone position information and energy status information. The two lasers that are close to each other charge the drone with low power, and the remaining laser charges the drone with high power.

[0029] (4) When information of 3 drones is obtained, the rotation of the three lasers is controlled by the obtained drone position information. The two lasers that are close to each other charge the drone with the lowest battery, and the remaining laser charges the drone with the second highest battery.

[0030] Beneficial effects of this invention:

[0031] This invention proposes an improved machine vision-based method for drone tracking and aiming, which can increase the energy level of transmission and improve heat dissipation, effectively reducing the size and weight of the transmitter and making it easier to install on a tower. It improves the speed and accuracy of tracking and aiming, and has significant theoretical and practical value for the development of laser-powered long-range drone technology.

[0032] Advantages of this invention:

[0033] (1) This invention breaks through the traditional thinking of laser wireless energy transmission, expanding the transmitter of a single laser to multiple lasers emitting simultaneously, and simplifies the transmitter turntable control system, using a single main control system to simultaneously control three turntables for synchronous tracking and aiming. Therefore, it can effectively improve the power level of the transmitter, and distributing the optical power to multiple transmitters also helps with the heat dissipation of the lasers. Sharing a single control system for the three turntables also helps reduce the cost of the turntable system and simplifies the control approach.

[0034] (2) This invention proposes an optimized charging scheme for multiple targets and multiple degrees of freedom, which solves the limitation that only a single UAV can be powered in actual operation. Moreover, multiple lasers can cover a wider charging space, and can achieve full-range coverage of multi-target power supply.

[0035] (3) The present invention proposes an improved machine vision positioning method, which compares and analyzes the advantages and disadvantages of the traditional monocular vision positioning detection method and the binocular vision positioning method, combines the advantages of the two methods, and makes up for the defects of the two vision positioning methods.

[0036] Monocular vision detection first searches for targets within the image. Once a target is found, its size in the image is estimated, and its distance is determined. The relative position of the lens and the target is then deduced from the target's position in the image. The advantage of this method is its low computational resource requirements and relatively simple system structure. The disadvantage is its relatively large error rate, making it impossible to guarantee high accuracy.

[0037] Binocular vision calculates the distance to a target by measuring the parallax between images captured by two cameras, and then processes the image using a target detection algorithm. The advantage of this method is its high measurement accuracy. However, it is relatively difficult to implement. The first challenge is the extremely large computational load, requiring very high performance from the computing unit, making the computational problems of binocular vision quite challenging in practical applications. Another challenge lies in the registration effect of binocular vision. Although stereo matching algorithms can be used to improve ranging accuracy, the principle of ranging itself requires minimal error between the two lenses. If each lens has an error of about 5%, the difficulty of subsequent algorithm calibration increases significantly, and determinism cannot be guaranteed. Furthermore, binocular positioning is more expensive than monocular positioning, and it also places higher demands on the installation positions of the two cameras.

[0038] This invention conducts comparative experiments on monocular and binocular visual positioning of UAVs under different flight attitudes and environments. It analyzes the accuracy of the two methods in different application scenarios and monitors the real-time frame rate of target detection during the experiments. Taking into account both frame rate and positioning accuracy, a backpropagation (BP) neural network model is built and trained to obtain the optimal positioning scheme in actual detection processes. This aims to improve the overall speed and accuracy of the tracking and aiming system.

[0039] This solves the technical problems that arise from the small area of ​​the laser spot and the photovoltaic array at the receiving end, the high speed of the UAV, and the high requirements for the tracking accuracy and real-time response speed of the transmitter turntable. Otherwise, the UAV will lose track of the target, resulting in energy loss and an inability to complete long-term flight. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall architecture of the multi-transmitter UAV tracking and aiming laser remote charging system according to a specific embodiment of the present invention.

[0041] Figure 2 This is a design diagram of the overall binocular vision positioning scheme according to a specific embodiment of the present invention;

[0042] Figure 3 This is a flowchart of a multi-target, multi-degree-of-freedom charging scheme according to a specific embodiment of the present invention. Detailed Implementation

[0043] A method for remote laser charging of multi-transmitter drones includes the use of three lasers operating in parallel at the transmitting end, with a small, lightweight air-cooling device for real-time heat dissipation. The system also establishes an integrated tracking and aiming device, which only needs to be installed on one of the turntables. This device uses binocular vision positioning to return the laser position information to the host, calculates the corresponding offset data, and transmits it to the two slave turntables. The slave turntables rotate according to commands, achieving simultaneous tracking of the three laser beams. This method ensures tracking and aiming accuracy while reducing cost and system complexity, and solves the current technical problem of insufficient laser power supply for drones.

[0044] To address the challenges of tracking and aiming due to the small area of ​​the photovoltaic receiver and the rapid movement of the UAV, a more accurate and reliable tracking and aiming scheme is proposed. This scheme combines binocular visual positioning with monocular camera image recognition positioning, establishing three image recognition methods: the first is a binocular visual positioning method using both cameras 1 and 2; the second is a traditional image recognition method using only camera 1, which calculates the relative size of the UAV in the captured image and converts it into its corresponding spatial position; the third is a traditional image recognition method using only camera 2, implemented in the same way as the second method. A BP neural network mathematical model is established for these three methods, and their tracking and aiming accuracy is analyzed under different scenarios and flight attitudes. Finally, appropriate weights are assigned to the three schemes to obtain the optimal tracking and aiming scheme for different scenarios and UAV flight attitudes.

[0045] To address the limitation of traditional wireless charging for drones, which can only charge a single target, an optimized algorithm is employed to enable wireless charging for multiple targets. This algorithm analyzes and calculates the battery level information transmitted from the drones to the transmitter when multiple targets are detected. It then controls the rotation of three lasers, prioritizing power to drones with lower battery levels. The algorithm analyzes three scenarios: charging a single drone, charging two drones, and charging three drones, ultimately forming a freely adjustable charging scheme. Specifically, it includes:

[0046] (1) By combining binocular visual positioning with monocular camera image recognition positioning technology, the position and energy status information of the UAV are obtained and stored in the computer.

[0047] (2) When information of one UAV is obtained, the rotation of three lasers is controlled by the obtained UAV location information, and the three lasers charge the UAV at the same time.

[0048] (3) When information of two drones is obtained, the three lasers are controlled to rotate by the obtained drone position information and energy status information. The two lasers that are closer together charge the drone with lower power, and the remaining laser charges the drone with higher power.

[0049] (4) When information of 3 drones is obtained, the rotation of the three lasers is controlled by the obtained drone position information. The two lasers that are closer together charge the drone with the lowest battery, and the remaining laser charges the drone with the second highest battery.

[0050] To address the issues of slow recognition speed and low frame rate in traditional image recognition systems, a lightweight neural network—YOLOv5-Lite—was used for model training. Simultaneously, the Jetson Nano development board replaced the traditional Raspberry Pi single-board computer used for image recognition, offering advantages over other AI computers in image recognition and object detection. To further improve model speed, the ONNXRuntime inference framework was used for optimization, and finally, the model was deployed to the powerful GPU—CUDA—to complete overall model optimization.

[0051] The technical solution of the invention will now be described in detail with reference to the accompanying drawings.

[0052] like Figure 1 The diagram shows the overall structure of a multi-transmitter UAV tracking and aiming laser remote charging system. It includes three parts: a laser system, a tracking and aiming system, and a photovoltaic receiving system. The tracking and aiming system includes cameras 1 and 2, an image acquisition card 3, a computer 4, a control circuit 5, and a servo motor 6. It is used to monitor the flight status of the UAV in real time and provide position information. Finally, it controls the rotation of the servo motor based on the position information to achieve tracking and aiming. The laser system includes lasers 7, 8, and 9. The three lasers are synchronously tracked by the turntable main control system, enabling the three lasers to be powered simultaneously. The photovoltaic receiving system includes the UAV 10 and a photovoltaic panel 11. The photovoltaic panel is mounted on the UAV's landing gear via a bracket and can be freely adjusted at a 30° angle to facilitate the photovoltaic array receiving laser light.

[0053] The following preparations are required before aiming:

[0054] (1) Take photos of the drone in various environments and flight attitudes in advance. The more photos there are, the more accurate the trained model will be in theory.

[0055] (2) Preprocess the captured images to unify their format, which facilitates feature extraction and model training.

[0056] (3) Build a YOLOv5-Lite lightweight neural network, perform multiple iterations of training on the preprocessed photo set, and set a validation set to estimate the accuracy of the trained model.

[0057] After preparation, two cameras are mounted on either side of the optical collimating lens to capture images of the drone in flight from different angles. The captured images are then transmitted to an image acquisition card, which in turn transmits the real-time image information to a computer. The computer calculates the drone's position based on two visual positioning theories. The main steps of binocular visual positioning are as follows:

[0058] Step 1: Real-time scene image acquisition: Binocular images are acquired through camera 1 and camera 2 to form a real-time scene image.

[0059] Step 2, Binocular camera calibration: Obtain the intrinsic and extrinsic parameters of cameras 1 and 2 respectively, generate a mapping model from image coordinates to world coordinates, and form a calibration model for the binocular cameras.

[0060] Step 3, Image Preprocessing: Using the acquired real-time scene images and the binocular camera calibration model, image preprocessing is performed on the acquired real-time scene images to improve image quality and accuracy, including: neighborhood averaging preprocessing and filtering preprocessing.

[0061] Step 4: Image feature point matching: The ORB algorithm is used to detect features in the images of the left and right cameras. The detected feature points are matched according to the matching algorithm, and the matched feature point pairs are recorded.

[0062] Step 5: Calculate the disparity based on the matched feature point pairs, and obtain the three-dimensional coordinates of the matched point pairs in the camera coordinate system according to the principle of binocular vision ranging.

[0063] Step 5: Calculate the camera pose using the PnP or ICP algorithm based on the 3D coordinates and image coordinates of the matching points.

[0064] Step 6: Determine the current position and attitude of the drone based on the relative positional relationship between the camera and the drone.

[0065] Step 7: Correct the positioning error by drawing on the ideas of visual SLAM and beacon recognition positioning.

[0066] Step 8: The UAV continuously moves and acquires position information in real time to obtain the global pose information of the UAV.

[0067] The computer transmits the acquired position information to the PLC control circuit, which controls the servo motors to rotate and achieve tracking. Since the three lasers are on the same horizontal plane, controlling their rotation only requires a slight horizontal angular offset. Because the drone's position relative to the central laser is known, and the horizontal distance between the two side lasers and the central laser can be set according to specific needs, the offset angle between the two side lasers and the central main control turntable laser can be calculated using the cosine theorem. This offset angle command is then transmitted to the PLC control circuits of the two side lasers, controlling the servo motors to rotate and achieve synchronous tracking of the three lasers. In actual operation, multiple drones may need to be charged simultaneously. If multiple drones (up to three) are detected in the camera's visual interface, the real-time battery information of the drones is transmitted to the transmitter using the wireless communication device between the drones and the transmitter. The transmitter then analyzes and calculates the battery levels of the multiple drones to obtain the optimal charging scheme, prioritizing charging the drones with low battery levels to ensure that all drones achieve the longest possible flight time.

[0068] In actual operation, before issuing a command to turn, the system will use a trained BP neural network to determine the appropriate visual positioning method based on the current flight status of the drone, the environment, and the current detection frame rate of the system, and adopt the best tracking and aiming scheme.

Claims

1. A method for remote laser charging of a multi-transmitter UAV, characterized in that: The method includes: using three lasers operating in parallel at the transmitting end to establish a multi-transmitter UAV tracking and laser remote charging system; simultaneously using a wind-cooling device to dissipate heat from the lasers in real time; the multi-transmitter UAV tracking and laser remote charging system establishes an integrated tracking device, installed on one of the turntables. The tracking device uses binocular vision positioning to return the laser's position information to the host, calculates the corresponding offset data, and transmits it to two slave turntables. The slave turntables rotate according to instructions to achieve simultaneous tracking of the three laser beams; the multi-transmitter UAV tracking and laser remote charging system includes: a laser system, a tracking system, and a photovoltaic receiving system; the laser system includes three lasers, which perform synchronous tracking according to the turntable's main control system, enabling simultaneous power supply to all three lasers; the laser's position information is returned to the host, the corresponding offset data is calculated and transmitted to the two slave turntables, and the slave turntables rotate according to instructions. The method for achieving simultaneous tracking of three laser beams by rotating the drone includes: the computer transmits the acquired position information to the PLC control circuit, which controls the servo motor to rotate and achieve tracking. Since the position information of the drone relative to the middle laser is known, and the horizontal distance between the two side lasers and the middle laser is set according to specific requirements, the offset angle between the two side lasers and the middle main control turntable laser is calculated according to the cosine theorem. The offset angle command is transmitted to the PLC control circuit of the two side lasers, which controls the servo motor to rotate, thereby achieving synchronous tracking of the three lasers. In actual operation, if more than one drone is detected in the camera's visual interface, the real-time battery information of the drone is transmitted to the transmitter using the wireless communication device between the drone and the transmitter. Then, the transmitter analyzes and calculates the battery level of each drone to obtain the optimal charging scheme, prioritizing charging the drone with low battery. The method for generating the optimal charging scheme is as follows: (1) By combining binocular visual positioning with monocular camera image recognition positioning technology, the position and energy status information of the UAV are obtained and stored in the computer; (2) When information about one UAV is obtained, the rotation of three lasers is controlled by the obtained UAV location information, so that the three lasers charge the UAV simultaneously. (3) When information of two drones is obtained, the three lasers are controlled to rotate by the obtained drone position information and energy status information. The two lasers that are close to each other charge the drone with low power, and the remaining laser charges the drone with high power. (4) When information of 3 drones is obtained, the rotation of the three lasers is controlled by the obtained drone position information. The two lasers that are close to each other charge the drone with the lowest battery, and the remaining laser charges the drone with the second highest battery.

2. The method for remote laser charging of a multi-transmitter UAV according to claim 1, characterized in that: The tracking and aiming system includes two cameras, an image acquisition card, a computer, control circuitry, and servo motors. It is used to monitor the drone's flight status in real time and provide position information. Finally, based on the position information, it controls the rotation of the servo motors to achieve tracking and aiming.

3. The method for remote laser charging of a multi-transmitter UAV according to claim 1, characterized in that: The photovoltaic receiving system includes a drone and photovoltaic panels. The photovoltaic panels are mounted on the drone's landing gear via brackets and can be freely adjusted at a 30° angle to facilitate the photovoltaic array receiving laser light.

4. The method for remote laser charging of a multi-transmitter UAV according to claim 1, characterized in that: The tracking and aiming device includes two cameras mounted on either side of an optical collimating lens to capture images of the drone in flight from different angles. The captured images are then transmitted to an image acquisition card, which transmits the real-time image information to a computer. The computer then calculates the drone's position information based on two visual positioning theories.

5. The method for remote laser charging of a multi-transmitter UAV according to claim 1, characterized in that: The tracking and aiming device uses a binocular vision positioning method including: Step 1: Real-time scene image acquisition: Binocular images are acquired through two cameras to form real-time images of the scene; Step 2, Binocular camera calibration: Obtain the intrinsic and extrinsic parameters of the two cameras respectively, generate a mapping model from image coordinates to world coordinates, and form a calibration model for the binocular camera; Step 3, Image Preprocessing: Using the acquired real-time scene images and the binocular camera calibration model, image preprocessing is performed on the acquired real-time scene images. Image preprocessing includes neighborhood averaging preprocessing and filtering preprocessing. Step 4, Image Feature Point Matching: The ORB algorithm is used to detect features in the images from the left and right cameras. The detected feature points are matched according to the matching algorithm, and the matched feature point pairs are recorded. Step 5: Calculate the disparity based on the matched feature point pairs, and obtain the three-dimensional coordinates of the matched point pairs in the camera coordinate system according to the principle of binocular vision ranging. Step 5: Calculate the camera pose using the PnP or ICP algorithm based on the 3D coordinates and image coordinates of the matching points; Step 6: Determine the current position and attitude of the drone based on the relative positional relationship between the camera and the drone; Step 7: Correct the positioning error by drawing on the ideas of visual SLAM and beacon recognition and positioning; Step 8: The UAV continuously moves and acquires position information in real time to obtain the global pose information of the UAV.

Images