Recycling method, system and equipment of unmanned aerial vehicle tower climbing anti-falling device and medium

Abstract

The invention provides a recovery method, system and equipment of an unmanned aerial vehicle tower climbing anti-falling device and a medium, and relates to the technical field of unmanned aerial vehicle inspection. The anti-falling device for tower climbing of the unmanned aerial vehicle comprises an unmanned aerial vehicle towing unit and a top hanging point unit. The method comprises the following steps that the unmanned aerial vehicle loaded with the unmanned aerial vehicle towing unit is controlled to fly to a preset butt joint position; determining a relative position relation between the top hanging point unit and the unmanned aerial vehicle through an image detection algorithm; according to the relative position relation between the top hanging point unit and the unmanned aerial vehicle, the flight path of the unmanned aerial vehicle is adjusted so as to achieve butt joint of the unmanned aerial vehicle towing unit and the top hanging point unit, and after butt joint is completed, the unmanned aerial vehicle is controlled to pull up the top hanging point unit and return. According to the technical scheme, the unmanned aerial vehicle tower climbing anti-falling device can be efficiently and safely recycled.

Description

Technical Field

[0001] The present application relates to the field of drone inspection technology, and more specifically, to a recovery method, system, equipment and medium for a drone tower climbing anti-fall device. Background Art

[0002] In industries such as power, communications, and construction, operations often need to be performed on tower tops or at heights. In order to prevent workers from falling during operations, fall prevention devices are usually installed. With the increase in the number of tower structures and the increase in the frequency of operations, how to efficiently and safely recycle these fall prevention devices has become a new challenge.

[0003] In the prior art, most of them rely on manual operation or simple mechanical devices to recycle these anti-fall devices. For example, the traditional manual recycling method requires workers to carry tools to climb to the top of the tower, manually disassemble and carry the anti-fall device down the tower. This method is not only inefficient, but also has extremely high safety risks. Although some recycling methods that use simple mechanical devices can reduce the burden on workers to a certain extent, their recycling efficiency and safety still need to be improved. In view of this, how to efficiently and safely recycle anti-fall devices has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present application is to provide a method, system, equipment and medium for recovering a drone tower climbing anti-fall device, which can efficiently and safely recover the drone tower climbing anti-fall device.

[0005] This application is implemented as follows:

[0006] In the first aspect, the present application provides a method for recovering a drone tower-climbing anti-fall device, wherein the drone tower-climbing anti-fall device includes a drone towing unit and a top hanging point unit, and the method includes the following steps: controlling a drone equipped with the drone towing unit to fly to a preset docking position; determining the relative position relationship between the top hanging point unit and the drone through an image detection algorithm; adjusting the flight path of the drone according to the relative position relationship between the top hanging point unit and the drone to achieve docking between the drone towing unit and the top hanging point unit, and controlling the drone to pull up the top hanging point unit and return after the docking is completed.

[0007] Further, based on the aforementioned scheme, the relative position relationship between the top hanging point unit and the drone is determined by an image detection algorithm, including: collecting a target image containing the top hanging point unit, and sending it into a target detection model for target detection to obtain an identification result of the top hanging point unit; performing ranging by a monocular or multi-camera carried by the drone to calculate the relative position relationship between the position of the top hanging point unit in the target image and the preset image center point, thereby determining the relative position relationship between the top hanging point unit and the drone.

[0008] Further, based on the aforementioned scheme, the target image containing the top hanging point unit is collected and sent to the target detection model for target detection to obtain the recognition result of the top hanging point unit, including: collecting the target image containing the top hanging point unit, and sending it to the target detection model for target detection, so that the target detection model obtains an output matrix through convolution calculation, and calculates the center coordinates of the border, the length of the border, the width of the border and the object confidence based on this, so as to select the top hanging point device, and eliminate redundant and overlapping frames through an improved algorithm to obtain the recognition result of the top hanging point unit.

[0009] Further, based on the aforementioned scheme, the method of eliminating redundant and overlapping frames through an improved algorithm includes: sorting the top hanging point devices selected by the frames by confidence, and calculating the intersection-and-union ratio between the frames; based on the calculated intersection-and-union ratio between the frames, deleting other frames whose intersection-and-union ratio with the highest confidence frame is greater than a preset threshold, and retaining the remaining frames as the final detection results.

[0010] Further, based on the aforementioned scheme, the distance measurement is performed by a monocular or multi-camera to calculate the relative position relationship between the position of the top hanging point unit in the target image and the center point of the preset image, including: using a binocular camera carried by a drone to obtain the position of the top hanging point unit in the two camera images; according to the position of the top hanging point unit in the two images, calculating the parallax and determining the offset direction of the top hanging point unit relative to the center point of the camera image; according to the internal parameters of the binocular camera and the calculated parallax, calculating the distance from the top hanging point unit to the binocular camera; according to the offset direction of the top hanging point unit relative to the center point of the camera image and the distance from the top hanging point unit to the binocular camera, determining the relative position relationship between the position of the top hanging point unit in the target image and the center point of the preset image.

[0011] Further, based on the above scheme, the distance measurement by a monocular or multi-camera includes: using a binocular camera to obtain the position information (X) of the center point of the hanging point in the left and right images R ,X T ), combined with the camera focal length f and the baseline B, the depth information Z of the target is calculated by the principle of similar triangles.

[0012] Further, based on the aforementioned scheme, the flight path of the UAV is adjusted according to the relative position relationship between the top hanging point unit and the UAV, including: controlling the UAV to fly to a position directly above the top hanging point unit according to the relative position relationship between the top hanging point unit and the UAV; measuring the distance between the top hanging point unit and the UAV in real time by a monocular or multi-camera carried by the UAV, and controlling the UAV to slowly lower its altitude at a certain speed according to the distance until the UAV towing unit and the top hanging point unit are successfully docked.

[0013] In the second aspect, the present application provides a recovery system for a drone tower-climbing anti-fall device, wherein the drone tower-climbing anti-fall device includes a drone towing unit and a top hanging point unit, and the system includes: a docking preparation module, configured to control a drone equipped with the drone towing unit to fly to a preset docking position; a position determination module, configured to determine the relative position relationship between the top hanging point unit and the drone through an image detection algorithm; a docking recovery module, configured to adjust the flight path of the drone according to the relative position relationship between the top hanging point unit and the drone, so as to achieve docking between the drone towing unit and the top hanging point unit, and control the drone to pull up the top hanging point unit and return after the docking is completed.

[0014] In a third aspect, the present application provides an electronic device, comprising a memory for storing one or more programs; a processor; when the one or more programs are executed by the processor, a method as described in any one of the first aspects is implemented.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method as described in any one of the above-mentioned first aspects.

[0016] Compared with the prior art, the present invention has at least the following advantages or beneficial effects:

[0017] This application proposes a method for recovering a drone tower climbing anti-fall device, which effectively solves the problems of low recovery efficiency, high safety risks, and low automation of anti-fall devices in the prior art through the combined application of drone technology, image detection algorithms, and precise flight control strategies, and achieves efficient, safe, and automated recovery of anti-fall devices. Specifically, it includes avoiding the high risk of manual climbing through remote operation of drones; improving the accuracy and efficiency of docking through image detection algorithms and flight control algorithms; and reducing manual intervention and improving overall operational efficiency through automated pull-up and return processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is a schematic diagram of the structure of a drone tower climbing anti-falling device in one embodiment of the present application;

[0020] Figure 2This is a flow chart of an embodiment of a method for recovering a tower climbing anti-fall device for a drone in the present application;

[0021] Figure 3 This is a schematic diagram of the principle of using a binocular camera to perform distance measurement in one embodiment of the present application;

[0022] Figure 4 This is a structural block diagram of an embodiment of a recovery system of a UAV tower climbing anti-fall device of the present application;

[0023] Figure 5 A structural block diagram of an electronic device provided in an embodiment of the present application.

[0024] Icons: 110, UAV towing unit; 111, connecting rod; 112, towing installation box; 120, top hanging point unit; 121, hanging ring; 122, anti-fall connector; 130, towing rope; 140, angle steel; 210, docking preparation module; 220, position determination module; 230, docking recovery module; 310, processor; 320, memory; 330, communication interface. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0026] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0027] Example 1

[0028] In the power, communication, construction and other industries, with the widespread application of tower structures (such as transmission towers, communication towers, construction cranes, etc.), the safety issues of high-altitude operations have become increasingly prominent. In particular, the anti-fall device, as a key equipment to ensure the life safety of high-altitude workers, its installation and recovery efficiency and safety directly affect the overall efficiency and safety of the operation.

[0029] The current mainstream recovery methods for anti-fall devices mainly rely on manual or simple mechanical devices. In the manual recovery method, workers are required to carry tools to climb to the top of the tower, manually disassemble and carry the heavy anti-fall device down the tower. This not only consumes a lot of physical strength and time, and increases the labor intensity of the workers, but also high-altitude operations themselves have extremely high safety risks, such as falling, electric shock and other dangers. In the recovery method with the help of simple mechanical devices, although the burden on the workers is reduced to a certain extent, due to the limited automation and intelligence level of the mechanical devices, the docking process is often not accurately controlled, resulting in low recovery efficiency, and it is easy to fail in complex environments, affecting the safety of recovery.

[0030] To this end, the present application provides a method for recovering a drone tower climbing anti-fall device, which utilizes the drone's high-altitude operation capability and precise control capability, combined with image detection algorithms and automated control technology, to achieve efficient and safe recovery of the anti-fall device.

[0031] It should be noted that in the technical solution of the present application, the drone tower climbing anti-fall device includes a drone towing unit 110 and a top hanging point unit 120. Among them, the top hanging point unit 120 is used to be mounted on the angle steel 140 of the tower structure and to set the traction rope 130. At the same time, after the drone is equipped with the drone towing unit 110, the drone towing unit 110 can be docked with the top hanging point unit 120, and the top hanging point unit 120 can be detached from the angle steel 140 after the docking is completed.

[0032] In order to facilitate the understanding of those skilled in the art, a specific example will be used here to illustrate the drone tower climbing anti-fall device.

[0033] like Figure 1 As shown, the drone towing unit 110 of this example includes a connecting rod 111 and a towing installation box 112, and the top hanging point unit 120 includes a hanging ring 121 and an anti-falling clamp 122. The drone towing unit 110 can be connected to the drone through the connecting rod 111, and docked with the hanging ring 121 through the towing installation box 112 (it can be locked by an electromagnetic lock after successful docking). The anti-falling clamp 122 can be clamped with the angle steel 140 (gravity self-locking), and when subjected to an upward traction force, the anti-falling clamp 122 can be separated from the angle steel 140.

[0034] See also Figure 2 The method for recovering a drone tower climbing anti-fall device comprises the following steps:

[0035] Step S101: Control the drone loaded with the drone towing unit 110 to fly to a preset docking position.

[0036] It should be noted that the drone towing unit 110 is a key component installed on the drone, and is designed with a docking mechanism that matches the top hanging point unit 120, which is used to firmly connect the two during the docking process. The unit may include a mechanical lock, electromagnetic adsorption or other advanced quick connection technology. The top hanging point unit 120 is part of the anti-fall device installed on the top of a tower or other high platform, serving as the docking target of the drone towing unit 110. It has an interface compatible with the drone towing unit 110 to ensure the accuracy and safety of the docking.

[0037] In the process of the drone loaded with the drone towing unit 110 being precisely controlled to fly to a preset docking position near the tower top, the GPS navigation system and advanced flight control algorithm of the drone can be relied upon. Specifically, in step S101, the drone can be programmed or remotely controlled to carry the drone towing unit 110 to a preset docking position. This preset docking position is usually determined based on the precise coordinates of the tower top or the target hanging point unit to ensure that the drone can approach to a close enough distance for subsequent docking operations. The flight path and altitude of the drone can be precisely controlled by an advanced flight control system (including GPS navigation, inertial navigation system, etc.) to ensure the stability and accuracy of the flight.

[0038] Step S102: Determine the relative position relationship between the top hanging point unit 120 and the UAV through an image detection algorithm.

[0039] After the drone approaches the preset docking position, in step S102, the camera on the drone is first used to capture the image of the top hanging point unit 120. Subsequently, the image detection algorithm processes and analyzes the captured image, identifies the specific position and posture of the top hanging point unit 120, and compares it with the current position of the drone, calculates the relative position relationship between the two (such as distance, angle, etc.), and provides a basis for subsequent flight adjustments. In addition, by accurately calculating the relative position relationship, the docking accuracy of the drone towing unit 110 and the top hanging point unit 120 can be greatly improved.

[0040] The application of the image detection algorithm enables the drone to obtain the position information of the top hanging point unit 120 in real time and compare and analyze it with its own position. This real-time feedback mechanism greatly improves the accuracy of docking and reduces the risk of docking failure caused by position deviation.

[0041] Step S103: According to the relative position relationship between the top hanging point unit 120 and the UAV, the flight path of the UAV is adjusted to achieve the docking of the UAV towing unit 110 and the top hanging point unit 120, and after the docking is completed, the UAV is controlled to pull up the top hanging point unit 120 and return.

[0042] In step S103, the flight path and attitude of the drone are dynamically adjusted mainly based on the relative position relationship information obtained in step S102. This may include adjusting parameters such as flight speed, altitude, heading angle, etc. to ensure that the drone towing unit 110 can accurately dock with the top hanging point unit 120. Once the docking is successful, the drone will activate the pulling mechanism (such as electromagnetic lock, mechanical lock, etc.) to firmly connect the top hanging point unit 120 to the towing unit. Subsequently, the drone control system controls the drone to pull up the top hanging point unit 120 and return safely according to the preset return path.

[0043] In summary, in the prior art, the work of recovering the anti-fall device mainly relies on manual climbing or simple mechanical devices. These recovery methods have significant disadvantages such as low efficiency, high safety risks, and low automation. In the above embodiments, the combined application of drone technology, image detection algorithms, and precise flight control strategies effectively solves the problems of low recovery efficiency, high safety risks, and low automation of anti-fall devices in the prior art, and achieves efficient, safe, and automated recovery of anti-fall devices. Specifically, it includes remote operation of drones to avoid the high risk of manual climbing; image detection algorithms and flight control algorithms to improve the accuracy and efficiency of docking; and automated pull-up and return processes to reduce manual intervention and improve overall operating efficiency.

[0044] Based on the aforementioned scheme, in one implementation of the present application, the relative position relationship between the top hanging point unit 120 and the drone is determined by an image detection algorithm, including: collecting a target image containing the top hanging point unit 120, and sending it into a target detection model for target detection to obtain an identification result of the top hanging point unit 120; performing ranging by a monocular or multi-camera carried by the drone to calculate the relative position relationship between the position of the top hanging point unit 120 in the target image and the preset image center point, thereby determining the relative position relationship between the top hanging point unit 120 and the drone.

[0045] In the above implementation, the process of determining the relative position relationship between the top hanging point unit 120 and the drone through the image detection algorithm is further refined and optimized (mainly including the target detection process, and the distance measurement and position relationship calculation process). This process not only involves target detection, but also combines distance measurement technology to provide more accurate and comprehensive location information.

[0046] The target detection process includes: first, using the high-definition camera (monocular or multi-camera) carried by the drone to collect the target image containing the top hanging point unit 120. Then the collected target image is sent to the pre-trained target detection model for processing. The target detection model can automatically identify the specific object in the image (i.e., the top hanging point unit 120) and give its specific position in the image (such as the bounding box coordinates). Through target detection, it can obtain the recognition result of the top hanging point unit 120, including its approximate position and shape information in the image.

[0047] The process of distance measurement and position relationship calculation includes: after obtaining the position of the top hanging point unit 120 in the image, the drone uses the onboard distance measurement sensor (which may be the depth estimation function of the monocular camera, or the stereo vision technology of the multi-eye camera) to measure the distance. The purpose of distance measurement is to obtain the actual distance between the top hanging point unit 120 and the drone. At the same time, the drone will also calculate the relative position relationship between the position of the top hanging point unit 120 in the target image and the preset image center point. The preset image center point can be the center of the image, or it can be other points set according to actual needs. By comparing the positions of the two, the offset direction and offset of the top hanging point unit 120 relative to the current viewing angle of the drone can be obtained. Finally, combining the actual distance and relative position relationship, the drone can determine the precise relative position relationship between the top hanging point unit 120 and itself.

[0048] In summary, in the above implementation, by combining target detection and ranging technology, the drone can more accurately determine the position of the top hanging point unit 120. This not only improves the success rate of docking, but also reduces repeated adjustments and flight time caused by position deviation. At the same time, the target detection model can identify top hanging point units 120 of different shapes, sizes and materials, thereby enhancing the system's adaptability to different environments. In addition, the precise relative position relationship information enables the drone to adjust the flight path and posture more quickly to achieve docking with the top hanging point unit 120. This shortens the operation time and improves the overall operation efficiency.

[0049] Furthermore, in an implementation of the present application, the target image containing the top hanging point unit 120 is collected and sent to the target detection model for target detection to obtain the recognition result of the top hanging point unit 120, including: collecting the target image containing the top hanging point unit 120, and sending it to the target detection model for target detection, so that the target detection model obtains an output matrix through convolution calculation, and calculates the center coordinates of the border, the length of the border, the width of the border and the object confidence based on this, so as to select the top hanging point device, and eliminate redundant and overlapping frames through an improved algorithm to obtain the recognition result of the top hanging point unit 120.

[0050] In the above implementation, the target detection model uses a multi-layer convolutional neural network to extract features from the input target image containing the top hanging point unit 120. These convolutional layers can automatically learn and extract useful features in the image, such as edges, corners, textures, etc. After multi-layer convolution and pooling operations, the target detection model generates an output matrix (or feature map). This matrix contains the probability information of whether there is a top hanging point unit 120 in each area of ​​the image, as well as the possible position and size of the top hanging point unit 120. Then based on the output matrix, the target detection model calculates the border center coordinates, border length, border width and object confidence of each potential top hanging point unit 120. These parameters together define the position and shape of the top hanging point unit 120 in the image.

[0051] Since the target detection model may generate multiple redundant or overlapping borders around the same top hanging point unit 120, these unnecessary borders are also removed by improving the algorithm in the above implementation. This process may involve strategies such as border merging (such as non-maximum suppression NMS algorithm) or border score sorting to ensure that only the most accurate and representative borders are retained in the end. After these processes, the final border will accurately select the top hanging point unit 120 and obtain the recognition result of the top hanging point unit 120.

[0052] In summary, the above implementation method uses a convolutional neural network to perform feature extraction and frame parameter calculation, and the target detection model can more accurately identify the top hanging point unit 120. At the same time, the accuracy and reliability of the recognition result are further improved by eliminating redundant and overlapping frames.

[0053] Furthermore, in one implementation of the present application, the elimination of redundant and overlapping frames through an improved algorithm includes: sorting the top hanging point devices selected by the frames by confidence, and calculating the intersection-and-union ratio between the frames; based on the calculated intersection-and-union ratio between the frames, deleting other frames whose intersection-and-union ratio with the highest confidence frame is greater than a preset threshold, and retaining the remaining frames as the final detection results.

[0054] In the above implementation, first, the borders of all the top hanging point devices selected by the preliminary box are sorted according to their confidence (that is, the probability that the model believes that there is a target object in the box). The higher the confidence of the box, the greater the possibility that it contains the real target object. Then, the intersection over union (IoU) between each box and other boxes is calculated. IoU is an indicator to measure the degree of overlap between two boxes, and its value ranges from 0 (no overlap) to 1 (complete overlap). By calculating IoU, it can be determined which boxes are redundant or overlapping. Then, according to the calculated IoU value, for each box with the highest confidence (that is, the first box in the current sorting), delete all other boxes whose IoU is greater than the preset threshold. This preset threshold is an empirical value, which is usually determined according to the actual application scenario and requirements. After the deletion operation, the retained box is part of the final detection result. Since some boxes are deleted, there may still be new boxes with the highest confidence in the remaining boxes. Therefore, it is necessary to repeat the above process (confidence sorting, calculating IoU, deleting redundant boxes) until there are no boxes that can be deleted. The remaining frame set is the final detection result of the top hanging point device.

[0055] In summary, the above implementation method significantly improves the accuracy and reliability of top hanging device detection by introducing confidence sorting and IoU calculation to remove redundant and overlapping borders, and optimizes the quality and clarity of the final detection results.

[0056] Based on the above scheme, in an implementation of the present application, the distance measurement is performed by a monocular or multi-camera to calculate the relative position relationship between the position of the top hanging point unit 120 in the target image and the preset image center point, including:

[0057] The binocular camera carried by the drone is used to obtain the position of the top hanging point unit 120 in the two camera images; based on the position of the top hanging point unit 120 in the two images, the parallax is calculated and the offset direction of the top hanging point unit 120 relative to the center point of the camera image is determined; based on the internal parameters of the binocular camera and the calculated parallax, the distance from the top hanging point unit 120 to the binocular camera is calculated; based on the offset direction of the top hanging point unit 120 relative to the center point of the camera image and the distance from the top hanging point unit 120 to the binocular camera, the relative position relationship between the position of the top hanging point unit 120 in the target image and the center point of the preset image is determined.

[0058] In the above implementation, the principle of binocular vision is combined to achieve accurate estimation of the three-dimensional position of the top hanging point unit 120. The specific implementation includes the following process: First, the binocular camera carried by the drone simultaneously captures the scene containing the top hanging point unit 120 and generates two slightly different images. The two images correspond to the left and right camera angles of the binocular camera respectively. Then, in the two camera images, the position of the top hanging point unit 120 is identified and extracted respectively. This is usually achieved by marking the bounding box or center point of the top hanging point unit 120 on the image. Then, according to the position difference of the top hanging point unit 120 in the two images (that is, the horizontal pixel difference of the corresponding points in the left and right images), the disparity is calculated. Disparity is a key parameter used to estimate the depth (that is, distance) of an object in binocular vision. Then, by comparing the position of the top hanging point unit 120 in the image with the position of the center point of the camera image, the offset direction of the top hanging point unit 120 relative to the center point of the camera image is determined. This includes offsets in both horizontal and vertical directions. Next, the actual distance from the top hanging point unit 120 to the binocular camera is calculated by triangulation using the internal parameters of the binocular camera (such as focal length, baseline distance, etc.) and the calculated parallax. This process usually involves a series of geometric calculations and mathematical transformations. Finally, combined with the offset direction of the top hanging point unit 120 relative to the center point of the camera image and the calculated actual distance, the relative position relationship between the position of the top hanging point unit 120 in the target image and the center point of the preset image can be determined. This includes the relative offset in the horizontal and vertical directions, as well as the actual three-dimensional distance information.

[0059] Among them, the use of binocular cameras for ranging can achieve accurate estimation of the three-dimensional position of the top hanging point unit 120. Because compared with monocular cameras, binocular cameras can provide more information about the depth of objects, thereby improving the accuracy and reliability of ranging. At the same time, the binocular vision system is less dependent on environmental factors such as lighting and weather, and can work in a variety of complex environments. This enables the drone to respond to various practical scenarios more flexibly when recovering the anti-fall device. And through the above-mentioned implementation method, the relative position relationship between the top hanging point unit 120 and the drone can be accurately calculated, and the drone will be able to more accurately adjust the flight path and posture to achieve precise docking with the top hanging point unit 120. This reduces errors and uncertainties in the docking process and improves work efficiency.

[0060] Based on the above scheme, in an implementation of the present application, the distance measurement by a monocular or multi-camera includes: using a binocular camera to obtain the position information (X) of the center point of the hanging point in the left and right images. R ,X T ), combined with the camera focal length f and the baseline B, the depth information Z of the target is calculated by the principle of similar triangles.

[0061] like Figure 3 As shown, P is the center point of the hanging point, O R With O T are the optical centers of the two cameras, points P' and P" are the imaging points of point P on the two cameras (the imaging plane of the camera is rotated and placed in front of the lens), f is the focal length of the camera, B is the baseline of the two camera centers, and Z is the depth information of the target. R is the horizontal position of the feature point in the left image, X T is the horizontal position of the corresponding feature point in the right image. Assume that the distance between points P' and P" is D, then: D = B-(XR-XT), according to the principle of similar triangles: D / B = (Zf) / Z. Then simplify the solution to get the value of Z. Among them, the focal length f and camera baseline B in the formula can be obtained through calibration.

[0062] Based on the above scheme, in one implementation of the present application, the flight path is adjusted according to the relative position relationship between the top hanging point unit 120 and the drone, and the strategy of real-time distance measurement and slow altitude reduction is combined to achieve efficient and safe docking between the drone and the top hanging point unit 120. Specifically, the flight path of the drone is adjusted according to the relative position relationship between the top hanging point unit 120 and the drone, including: controlling the drone to fly to the position directly above the top hanging point unit 120 according to the relative position relationship between the top hanging point unit 120 and the drone; measuring the distance between the top hanging point unit 120 and the drone in real time through a monocular or multi-camera carried by the drone, and controlling the drone to slowly reduce its altitude at a certain speed according to the distance until the drone towing unit 110 and the top hanging point unit 120 are successfully docked.

[0063] In the above implementation, after the relative position relationship between the top hanging point unit 120 and the drone is obtained, the drone is first controlled to fly to the position directly above the top hanging point unit 120 according to the calculated relative position relationship. This involves adjusting the heading, pitch and yaw angles of the drone to ensure that the drone can accurately reach the top of the target point. At the same time, the distance between the top hanging point unit 120 and the drone is measured in real time by the camera carried by the drone. Then, according to the distance information measured in real time, the drone is controlled to lower its altitude at a certain speed (usually slow). In this process, it is necessary to continuously monitor the distance information to ensure that the drone does not collide with the top hanging point unit 120 due to too fast altitude reduction. When the drone is lowered to a sufficiently close distance, the towing unit (such as a mechanical arm, a hook, etc.) of the drone is started to dock with the top hanging point unit 120. During the docking process, it may be necessary to fine-tune the position and attitude of the drone to ensure the accuracy and stability of the docking. Finally, after the docking is completed, the camera or other sensors are used to confirm whether the docking is successful. If successful, the drone can continue to perform subsequent tasks; if it fails, it is necessary to re-dock or take other remedial measures.

[0064] In short, by measuring the distance between the top hanging point unit 120 and the drone in real time and controlling the drone to slowly lower its altitude, the accuracy and stability of docking can be greatly improved. This helps to reduce failures and losses caused by inaccurate docking. In addition, during the docking process, continuously monitoring the distance information and controlling the speed and attitude of the drone can ensure a safe distance between the drone and the top hanging point unit 120 to avoid accidents such as collisions. At the same time, through automated and intelligent flight path adjustment, the drone can complete the docking task faster, thereby improving overall operational efficiency.

[0065] In order to make the technical personnel in this field understand the present application more intuitively, a specific example will be used here to illustrate the recovery method of the drone tower climbing anti-fall device proposed in the present application. It can be understood that the example is only for the technical personnel in this field to understand the present application more intuitively, and does not limit the implementation of the present application to adopt the same measures.

[0066] First, it is necessary to control the drone equipped with the drone towing unit 110 to fly to the preset docking position. After the drone flies to the preset docking position, it starts to collect target images including the top hanging point unit 120, and inputs the collected target images into the yolov5 target detection model to detect the target images. Among them, the target detection model is trained in advance. During the training process, it is necessary to obtain a series of image data containing the top hanging point unit 120, and use labelimg to annotate the data of the top hanging point unit 120 in the image, and then use the pre-trained model and multi-scale training method for training. After the training is completed, the obtained target detection model can be deployed to the onboard box of the drone, and the image obtained by the camera is input into the model for reasoning to obtain the position information of the top hanging point unit 120 in the image, where the position information includes the target x, y coordinates, wight, hight, etc.

[0067] Specifically, the target detection process includes: after the target image is input into the target detection model, the YOLOV5 neural network divides the input data set into 7×7×30 feature maps, and obtains the output matrix through neural network convolution calculation. The neural network analysis matrix is ​​used to calculate the coordinates of the center of the frame, the length of the frame, the width of the frame and the confidence of the object. After the YOLOV5 neural network is executed, the top hanging point unit 120 is selected according to the above coordinates to obtain a preliminary detection result. Then, the improved algorithm can be used to remove redundancy and overlapping frames from the preliminary detection results. For example, the frames output by the target detection model can be sorted from large to small according to the confidence, and the sorted results are placed in a list. Then take out the first-ranked frame from the list, and use it to calculate the intersection and union ratio with other frames. Once there are other frames in the list whose intersection and union ratio is greater than the threshold, these frames will be deleted, and the first-ranked frame will be taken out from the sorted list and placed in the candidate frame list. Until the sorted list is empty, the feature objects selected by the remaining frames are retained to achieve the purpose of eliminating redundancy. Finally, the preliminary detection results are identified and screened again using the improved YOLOV5 algorithm mentioned above to obtain the final detection results.

[0068] Next, the relative position relationship between the top hanging point unit 120 and the drone can be determined by an image detection algorithm. Exemplarily, this can be implemented by the following steps:

[0069] (1) Detect the position of the top hanging point unit 120

[0070] First, a binocular camera is used to obtain the position information of the top hanging point unit 120 in the two images. Some computer vision methods, such as feature point matching, stereo vision and other technologies, can be used to detect the position of the top hanging point unit 120 in the left and right images.

[0071] Specifically, key feature points, such as corner points, edge points or descriptors, can be detected in the images of the left and right cameras. Descriptors are extracted from these feature points, such as SIFT (Scale Invariant Feature Transform) or ORB (Oriented FAST and Rotated BRIEF) and other algorithms. Then, the feature points in the left and right images are matched. The matching goal is to find which feature point in the right image corresponds to the feature point in the left image. The matching algorithm can use a descriptor-based algorithm, such as the nearest neighbor matching algorithm, based on a distance threshold or an optimization algorithm (such as the RANSAC algorithm) to eliminate false matches.

[0072] (2) Calculating parallax

[0073] Once the feature points are successfully matched, the difference in their horizontal positions in the left and right images, i.e., disparity, can be calculated. Disparity can be calculated by the following formula: Disparity = x_l-x_r, where (x_l) is the horizontal position of the feature point in the left image, and (x_r) is the horizontal position of the corresponding feature point in the right image. Then the disparity to depth conversion can be performed: the disparity is inversely proportional to the depth of the object, and the depth can be calculated by the internal and external parameters of the stereo vision system and the baseline length between the cameras. The calculation formula for depth Depth can be: Depth = f×B / Disparity. Where f is the focal length of the camera and B is the baseline distance between the cameras.

[0074] (3) Calculate distance

[0075] Using the parallax information and the camera's internal parameter matrix, the distance from the top hanging point unit 120 to the camera can be calculated using triangulation. This can be calculated using the camera's focal length, baseline length and other parameters.

[0076] (4) Calculation direction

[0077] According to the position information of the target in the image (the target object is the anti-falling clip 122, not the hanging ring 121. During the deep learning training process, the distance judgment is mainly based on the anti-falling clip 122. Since the hanging ring 121 is of fixed size, it can be solved by the relative position of the anti-falling clip 122. This design is mainly to facilitate the buckle of the towing installation box 112 to face the hanging ring 121. During the towing downward process, the hanging ring 121 is slightly moved as the target until the buckle of the towing installation box 112 reaches the hanging ring 121 to trigger the lock tongue) the offset direction of the target relative to the center of the image can be determined.

[0078] The position information of the top hanging point unit 120 in the target image is determined by the image detection algorithm, and the preset image center point coordinates are taken as (x, y). Based on the image center point as the origin, the relative position of the UAV and the top hanging point unit 120 is calculated. Then the UAV moves toward the top hanging point unit 120, and the center of the top hanging point unit 120 is made to coincide with the center of the target image, so that the UAV is located vertically above the top hanging point unit 120.

[0079] Specifically, assume that the preset image center point (x, y) is a point in the vertical direction of the drone, and (x1, y1) is the pixel coordinate of the center point of the top hanging point unit 120 in the target image. To move (x1, y1) in the (x, y) direction and make them coincide, the direction and distance judgment process is: calculate the value of (x1-x), if the value is positive, the top hanging point unit 120 is on the right side of the drone, and the drone needs to move horizontally to the right; otherwise, if the value is negative, the top hanging point unit 120 is on the left side of the drone, and the drone needs to move to the left. When (y1-y) is positive, the top hanging point unit 120 is behind the drone, and the drone needs to move backward; otherwise, (y1-y) is negative, the top hanging point unit 120 is in front of the drone, and the drone needs to move forward.

[0080] Finally, by comprehensively utilizing the above information, the distance and direction from the top hanging point unit 120 to the center of the image can be accurately calculated, providing important positioning information for the UAV flight. After the UAV reaches directly above the top hanging point unit 120, the UAV is controlled to slowly lower its altitude at a certain speed, and the distance between the target and the UAV is measured in real time by binocular ranging until the recovery device feeds back the capture signal.

[0081] like Figure 2 As shown, if the drone tower climbing anti-fall device includes two large modules, namely the drone towing unit 110 and the top hanging point unit 120. The drone towing unit 110 can be further composed of an edge airborne box (the edge airborne box will execute the program processing task to realize the recovery method of the drone tower climbing anti-fall device shown in this application), a connecting rod 111, and a bottom towing installation box 112. The edge airborne box contains a monocular or binocular camera, a communication module, a recognition algorithm module, and a control algorithm module. Among them, the connecting rod 111 adopts a flexible joint structure, that is, the flexible joint is a connector composed of multiple sections of straight tubes of ultra-light materials. The bottom towing installation box 112 is a fork structure (eight-shaped structure), which contains a towing installation box 112 control module, a posture sensor, a lock tongue sensing control module, a lock tongue drive motor, a metal detection closure sensing module, and a closure sensing light. The top hanging point unit 120 is composed of an anti-falling clamping piece 122 and a hanging ring 121 , and the hanging ring 121 is vertically centered with the anti-falling clamping piece 122 .

[0082] Therefore, when the drone flies to the preset docking position, the edge airborne box starts the automatic identification and control process:

[0083] (1) The edge airborne box sends a real-time detection command to the lock tongue sensing opening and closing module in the trailer installation box 112 through the serial port. At this time, the trailer device enters the real-time sensing logic of whether the hanging ring 121 is closed. If it is not closed, the command to allow the airborne box to automatically control the drone is sent to the airborne box through the control board.

[0084] (2) Then, the onboard box enters the image recognition process, determines the relative position relationship between the top hanging point unit 120 and the drone through the image detection algorithm, and sends the control command to the flight control system of the drone based on the relative position relationship. The flight control system controls the drone to move horizontally in the center. After the drone is centered, it is necessary to adjust the orientation according to the top hanging point unit 120 below, otherwise the drone towing unit 110 and the hanging ring 121 may not be aligned.

[0085] (3) Until the drone has flown directly above the top hanging point unit 120, it starts the descent operation. During the descent process, the perception of another attitude sensor in the drone towing unit 110 will be triggered. The reason for the need for this perception data is that the drone towing unit 110 and the drone are flexibly connected (in this solution, if the rigid connection causes the drone towing unit 110 to touch any object, the drone will think it is on the ground, thereby stopping the blades from rotating and triggering the explosion). The flexibly connected drone towing unit 110 will swing in the air, and it cannot automatically dock with the hanging ring 121 when swinging, so it is necessary to wait for the sensor to sense that the drone towing unit 110 has stopped swinging.

[0086] (4) When the towing device stops swinging, it notifies the edge airborne box to control the drone to descend. When it descends to the hanging ring 121 and triggers the closing sensor, the lock tongue is immediately locked, and the closing state is synchronously transmitted to the airborne box, which sends a command to the drone to stop moving downward.

[0087] (5) After the edge airborne box sends a command to stop moving downward, the closed state is fed back to the remote control, and the light bar directly above the fork-shaped trailer installation box 112 is lit. At this time, the remote control or ground station will display the closed state, and the light bar on the trailer installation box 112 can also be observed with the naked eye. At this time, it is a double closure confirmation.

[0088] (6) After the edge airborne box sends the stop downward movement command, there is a delay of 100ms. If no personnel take over the control of the drone to pull up the hanging point device, the edge airborne box controls the drone to take off automatically, pulls up the top hanging point unit 120 and flies to the return point. At this time, the closed loop of the entire link is completed.

[0089] It should be noted that in the above example of the anti-fall device for climbing a tower, the reason why the drone and the towing device are designed to be flexibly connected is to prevent the drone from exploding, because when the drone towing unit 110 descends and approaches the hanging ring 121, there must be contact. If it is a rigid connection, the drone will think that it has touched the ground, and thus stop the rotation of the blades, resulting in an explosion accident. In addition, the towing installation box 112 is designed as a fork type and the hanging ring 121 is set as an arc, mainly because there is a slight shake during the descent of the drone. The towing installation box 112 is designed as a lower opening fork type, and the top of the hanging ring 121 is designed as an elliptical arc. During the descent of the drone, the drone towing unit 110 will swing slightly with the flexible joint. When the drone towing unit 110 is just stuck on the hanging ring 121, it is a normal state, but when shaking occurs, the drone may move abnormally for a certain distance, and the flexible joint just guarantees the redundancy of the distance, so that the towing installation box 112 and the hanging ring 121 present a guided locking closure.

[0090] Example 2

[0091] See also Figure 4 The embodiment of the present application provides a recovery system for a drone tower climbing anti-fall device, wherein the drone tower climbing anti-fall device includes a drone towing unit 110 and a top hanging point unit 120, and the system includes: a docking preparation module 210, configured to control a drone loaded with the drone towing unit 110 to fly to a preset docking position; a position determination module 220, configured to determine the relative position relationship between the top hanging point unit 120 and the drone through an image detection algorithm; a docking recovery module 230, configured to adjust the flight path of the drone according to the relative position relationship between the top hanging point unit 120 and the drone, so as to achieve docking between the drone towing unit 110 and the top hanging point unit 120, and control the drone to pull up the top hanging point unit 120 and return after the docking is completed.

[0092] For the specific implementation process of the above system, please refer to the recovery method of the drone tower climbing anti-fall device provided in Example 1, which will not be repeated here.

[0093] Example 3

[0094] See also Figure 5, the embodiment of the present application provides an electronic device, which includes at least one processor 310 and at least one memory 320; wherein the processor 310 and the memory 320 are directly connected to each other, or communicate with each other through a communication interface 330, or are electrically connected through one or more communication buses or signal lines to achieve data transmission or interaction; the memory 320 stores program instructions that can be executed by the processor 310, and the processor 310 calls the program instructions to execute a method for recovering a drone tower climbing anti-fall device. For example, the following is realized:

[0095] The unmanned aerial vehicle tower climbing anti-fall device includes a unmanned aerial vehicle towing unit 110 and a top hanging point unit 120, and the recovery method of the unmanned aerial vehicle tower climbing anti-fall device includes the following steps: controlling the unmanned aerial vehicle loaded with the unmanned aerial vehicle towing unit 110 to fly to a preset docking position; determining the relative position relationship between the top hanging point unit 120 and the unmanned aerial vehicle through an image detection algorithm; adjusting the flight path of the unmanned aerial vehicle according to the relative position relationship between the top hanging point unit 120 and the unmanned aerial vehicle to achieve docking between the unmanned aerial vehicle towing unit 110 and the top hanging point unit 120, and controlling the unmanned aerial vehicle to pull up the top hanging point unit 120 and return after the docking is completed.

[0096] Among them, the memory 320 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0097] The processor 310 may be an integrated circuit chip with signal processing capability. The processor 310 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0098] Understandably, Figure 5 The structure shown is for illustration only. The electronic device may also include Figure 5 More or fewer components as shown, or with Figure 5 Different configurations are shown. Figure 5 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0099] Example 4

[0100] The present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor 310, implements a method for recovering a drone tower climbing anti-fall device. For example, the following is implemented:

[0101] The unmanned aerial vehicle tower climbing anti-fall device includes a unmanned aerial vehicle towing unit 110 and a top hanging point unit 120, and the recovery method of the unmanned aerial vehicle tower climbing anti-fall device includes the following steps: controlling the unmanned aerial vehicle loaded with the unmanned aerial vehicle towing unit 110 to fly to a preset docking position; determining the relative position relationship between the top hanging point unit 120 and the unmanned aerial vehicle through an image detection algorithm; adjusting the flight path of the unmanned aerial vehicle according to the relative position relationship between the top hanging point unit 120 and the unmanned aerial vehicle to achieve docking between the unmanned aerial vehicle towing unit 110 and the top hanging point unit 120, and controlling the unmanned aerial vehicle to pull up the top hanging point unit 120 and return after the docking is completed.

[0102] If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0103] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for recovering a drone tower climbing anti-fall device, characterized in that: The UAV tower climbing anti-fall device comprises a UAV towing unit and a top hanging point unit, and the method comprises the following steps: Control the UAV equipped with the UAV trailer unit to fly to a preset docking position; Determine the relative position relationship between the top hanging point unit and the UAV through image detection algorithm; According to the relative position relationship between the top hanging point unit and the UAV, the flight path of the UAV is adjusted to achieve the docking of the UAV towing unit and the top hanging point unit, and after the docking is completed, the UAV is controlled to pull up the top hanging point unit and return.

2. The method according to claim 1, characterized in that Determining the relative position relationship between the top hanging point unit and the UAV by using an image detection algorithm includes: Collect the target image containing the top hanging point unit and send it to the target detection model for target detection to obtain the recognition result of the top hanging point unit; The distance measurement is performed through the monocular or multi-camera carried by the UAV to calculate the relative position relationship between the position of the top hanging point unit in the target image and the center point of the preset image, thereby determining the relative position relationship between the top hanging point unit and the UAV.

3. The method according to claim 2, characterized in that The acquisition includes a target image of the top hanging point unit and is sent to a target detection model for target detection to obtain a recognition result of the top hanging point unit, including: A target image containing a top hanging point unit is collected and sent to a target detection model for target detection, so that the target detection model obtains an output matrix through convolution calculation, and calculates the center coordinates of the border, border length, border width and object confidence based on this, so as to select the top hanging point device, and eliminate redundant and overlapping frames through an improved algorithm to obtain the recognition result of the top hanging point unit.

4. The method according to claim 3, characterized in that The method of removing redundant and overlapping frames by improving the algorithm includes: Confidence ranking of the top hanging point devices selected by the boxes is performed, and the intersection-and-union ratio between the boxes is calculated; According to the calculated IoU ratio between each box, other boxes whose IoU ratio with the highest confidence box is greater than the preset threshold are deleted, and the remaining boxes are retained as the final detection results.

5. The method according to claim 2, characterized in that: The distance measurement is performed by a monocular or multi-camera to calculate the relative position relationship between the position of the top hanging point unit in the target image and the center point of the preset image, including: Use the binocular camera carried by the drone to obtain the position of the top hanging point unit in the two camera images; According to the position of the top hanging point unit in the two images, the parallax is calculated and the offset direction of the top hanging point unit relative to the center point of the camera image is determined; According to the internal parameters of the binocular camera and the calculated parallax, the distance from the top hanging point unit to the binocular camera is calculated; According to the offset direction of the top hanging point unit relative to the center point of the camera image and the distance from the top hanging point unit to the binocular camera, the relative position relationship between the position of the top hanging point unit in the target image and the preset image center point is determined.

6. The method according to claim 2, characterized in that The distance measurement by a monocular or multi-camera includes: using a binocular camera to obtain the position information (X) of the center point of the hanging point in the left and right images. R ,X T ), combined with the camera focal length f and the baseline B, the depth information Z of the target is calculated by the principle of similar triangles.

7. The method according to claim 1, characterized in that The step of adjusting the flight path of the UAV according to the relative position relationship between the top hanging point unit and the UAV includes: According to the relative position relationship between the top hanging point unit and the UAV, the UAV is controlled to fly to a position directly above the top hanging point unit; The distance between the top hanging point unit and the drone is measured in real time through the monocular or multi-camera carried by the drone, and the drone is controlled to slowly lower its altitude at a certain speed according to the distance until the drone towing unit and the top hanging point unit are successfully docked.

8. A recovery system for a drone tower climbing anti-fall device, characterized in that: The UAV tower climbing anti-fall device includes a UAV towing unit and a top hanging point unit, and the system includes: The docking preparation module is configured to: control the UAV equipped with the UAV trailer unit to fly to a preset docking position; The position determination module is configured to: determine the relative position relationship between the top hanging point unit and the UAV through an image detection algorithm; The docking and recovery module is configured to: adjust the flight path of the UAV according to the relative position relationship between the top hanging point unit and the UAV to achieve the docking of the UAV towing unit and the top hanging point unit, and control the UAV to pull up the top hanging point unit and return after the docking is completed.

9. An electronic device, characterized in that: include: A memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images