Power patrol unmanned aerial vehicle fixed-point landing method, system, equipment and medium

Abstract

The invention relates to the technical field of electric power systems, and discloses a fixed-point landing method, system, equipment and medium for an electric power patrol unmanned aerial vehicle, and the method comprises the steps: collecting color image data of a target landing platform corresponding to an unmanned aerial vehicle end in real time, and fusing visual positioning information and radar positioning information through a Kalman filtering algorithm, based on a PID control algorithm, the current flight information of the unmanned aerial vehicle end is adaptively adjusted according to the fused positioning information and the current flight information of the unmanned aerial vehicle end, a control instruction of the unmanned aerial vehicle end is determined and executed according to the flight information adjustment amount of the unmanned aerial vehicle end, and the current flight information of the unmanned aerial vehicle end is updated. And the flight information is continuously updated in the landing process until the unmanned aerial vehicle end lands on the target landing platform, so that the precision and reliability of fixed-point landing of the unmanned aerial vehicle are improved, and the unmanned aerial vehicle landing requirements in a complex environment are met.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, system, equipment and medium for fixed-point landing of a power patrol drone. Background Art

[0002] With the rapid development of drone technology, drones are increasingly used in fields such as electricity. However, safe and accurate landing of drones has always been a key problem in the industry. Traditional landing methods based on visual recognition are highly dependent on light conditions. In environments with strong light, weak light, or drastic light changes, their recognition accuracy will drop significantly, making it difficult to ensure the accuracy of landing. However, the accuracy of landing based solely on GPS positioning is severely limited by the accuracy of satellite signals and signal obstruction. In signal-blocked areas such as urban high-rise buildings and mountainous areas, the landing deviation may reach several meters or even greater, which results in poor accuracy and reliability of drone fixed-point landing, making it difficult to meet the landing needs of drones in complex environments. Summary of the invention

[0003] In view of this, the present invention provides a method, system, equipment and medium for fixed-point landing of power patrol UAVs, which solves the technical problems that the accuracy and reliability of fixed-point landing of UAVs are poor and it is difficult to cope with the landing needs of UAVs in complex environments.

[0004] The first aspect of the present invention provides a method for fixed-point landing of a power patrol UAV, which is applied to a UAV end and includes:

[0005] Collecting color image data of the target landing platform corresponding to the UAV end in real time;

[0006] Positioning the target landing platform according to the color image data to obtain visual positioning information;

[0007] Performing point cloud scanning on the target landing platform by using a laser radar to obtain radar positioning information of the target landing platform;

[0008] Using a Kalman filter algorithm to fuse the visual positioning information and the radar positioning information to obtain fused positioning information;

[0009] Based on the PID control algorithm, the current flight information of the UAV is adaptively adjusted according to the fused positioning information and the current flight information of the UAV to obtain the flight information adjustment amount of the UAV;

[0010] Determine the control instruction of the drone end according to the flight information adjustment amount of the drone end, execute the control instruction of the drone end, and update the current flight information of the drone end;

[0011] Based on the updated current flight information of the drone end, the PID control algorithm is re-executed to determine the step of adjusting the flight information of the drone end according to the fused positioning information and the current flight information of the drone end until the drone end lands on the target landing platform.

[0012] Preferably, the method further comprises preprocessing the color image data;

[0013] The preprocessing of the color image data comprises:

[0014] Performing grayscale processing on the color image data to obtain a grayscale image;

[0015] Determine the grayscale value of each pixel in the grayscale image by weighted average method;

[0016] Performing noise filtering on the grayscale image by Gaussian filtering and median filtering according to the grayscale value of each pixel in the grayscale image;

[0017] The Sobel algorithm is used to enhance the edge features of the grayscale image after noise filtering.

[0018] Preferably, positioning the target landing platform according to the color image data to obtain visual positioning information includes:

[0019] Identify a bounding box of the target landing platform in the preprocessed color image data based on an image recognition neural network; wherein the image recognition neural network is obtained by pre-training a historical training sample set based on a deep convolutional neural network, and the historical training sample set includes historical image samples of the target landing platform and bounding boxes corresponding to the historical image samples;

[0020] Extracting bounding box features of the bounding box, wherein the bounding box features include bounding corner points and a center point;

[0021] Based on the camera intrinsic parameter matrix, the coordinates of the bounding box feature in the two-dimensional coordinate system are converted into three-dimensional coordinates in the camera coordinate system;

[0022] Based on the camera extrinsic matrix, the three-dimensional coordinates of the bounding box features in the camera coordinate system are converted to the world coordinate system, and the visual positioning information is determined according to the three-dimensional coordinates of the bounding box features in the world coordinate system.

[0023] Preferably, performing point cloud scanning on the target landing platform by using a laser radar to obtain radar positioning information of the target landing platform includes:

[0024] Scanning the target landing platform by a laser radar to obtain three-dimensional point cloud data of the target landing platform;

[0025] Performing noise filtering on the three-dimensional point cloud data using a statistical filtering algorithm;

[0026] The voxel filtering algorithm is used to downsample the three-dimensional point cloud data after noise filtering;

[0027] Performing clustering processing on the downsampled three-dimensional point cloud data based on a region growing algorithm to extract point cloud region data of the target landing platform;

[0028] Determine a normal equation according to the point cloud area data of the target landing platform, and solve the normal equation using the least square method to obtain the normal vector and center point coordinates of the target landing platform;

[0029] The radar positioning information of the target landing platform relative to the UAV end is determined according to the normal vector and the center point coordinates of the target landing platform, and the radar positioning information includes position information and attitude information.

[0030] Preferably, the using of a Kalman filter algorithm to fuse the visual positioning information and the radar positioning information to obtain fused positioning information includes:

[0031] Constructing a visual observation matrix and a radar observation matrix according to the visual positioning information and the radar positioning information;

[0032] Constructing a visual observation equation according to the visual observation matrix and the system state vector;

[0033] Constructing a radar observation equation according to the radar observation matrix and the system state vector;

[0034] Use the state equation to predict the current state and covariance matrix;

[0035] Update the Kalman gain by using the visual observation equation and the radar observation equation;

[0036] The Kalman gain is used to update the state and covariance matrix at the current moment until the update iteration converges, and the state after the update iteration converges is output as the fused positioning information.

[0037] Preferably, the fused positioning information includes the position information of the target landing platform; the flight information adjustment amount includes the position adjustment amount and the angle adjustment amount;

[0038] The method based on the PID control algorithm, adaptively adjusting the current flight information of the drone end according to the fused positioning information and the current flight information of the drone end, and obtaining the flight information adjustment amount of the drone end, includes:

[0039] According to the position deviation between the current position of the drone and the position information of the target landing platform;

[0040] Adaptively adjusting the current position of the drone end through a PID control algorithm according to the position deviation to obtain a position adjustment amount; wherein the position adjustment amount includes a horizontal position adjustment amount and a vertical position adjustment amount;

[0041] According to the angle deviation between the current posture of the drone and the expected posture, the current angle deviation of the drone is adaptively adjusted through a PID control algorithm to obtain an angle adjustment amount; the angle adjustment amount includes an Euler angle adjustment amount, a pitch angle adjustment amount, and a roll angle adjustment amount.

[0042] Preferably, a plurality of pressure sensors are evenly distributed on the target landing platform, and the method further comprises:

[0043] Monitoring the pressure information of the drone after landing on the target landing platform according to the plurality of pressure sensors;

[0044] When the pressure information monitored by the plurality of pressure sensors reaches or exceeds a preset pressure threshold, it is determined that the landing operation of the drone end to land on the target landing platform is successful, and a landing success signal is generated and sent to the drone end;

[0045] When the drone receives the landing success signal, the flight operation of the drone is turned off, so that the drone stops on the target landing platform.

[0046] In a second aspect, the present invention further provides a fixed-point landing system for an electric patrol UAV, which is applied to a UAV end, wherein the UAV end is provided with a visual perception module, a visual positioning module, a laser radar ranging module, a data fusion module and a flight control module;

[0047] The visual perception module is used to collect color image data of the target landing platform corresponding to the drone end in real time;

[0048] The visual positioning module is used to locate the target landing platform according to the color image data to obtain visual positioning information;

[0049] The laser radar ranging module is used to perform point cloud scanning on the target landing platform through the laser radar to obtain radar positioning information of the target landing platform;

[0050] The data fusion module is used to fuse the visual positioning information and the radar positioning information using a Kalman filter algorithm to obtain fused positioning information;

[0051] The flight control module is used to adaptively adjust the current flight information of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end, so as to obtain the flight information adjustment amount of the drone end; wherein the current flight information of the drone end includes the current flight positioning information and the current flight attitude information; it is also used to determine the control instructions of the drone end according to the flight information adjustment amount of the drone end, and execute the control instructions of the drone end to update the current flight information of the drone end; it is also used to re-execute the process of determining the flight information adjustment amount of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end based on the updated current flight information of the drone end, until the drone end lands on the target landing platform.

[0052] In a third aspect, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for fixed-point landing of a power patrol drone as described in the first aspect.

[0053] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the steps of the method for fixed-point landing of a power inspection UAV as described in the first aspect.

[0054] It can be seen from the above technical scheme that the present invention obtains fused positioning information by collecting color image data of the target landing platform corresponding to the drone end in real time, and uses the Kalman filter algorithm to fuse the visual positioning information and the radar positioning information. Based on the PID control algorithm, the current flight information of the drone end is adaptively adjusted according to the fused positioning information and the current flight information of the drone end, the control instructions of the drone end are determined and executed according to the adjustment amount of the flight information of the drone end, the current flight information of the drone end is updated, and the flight information is continuously updated during the landing process until the drone end lands on the target landing platform, thereby improving the accuracy and reliability of the drone's fixed-point landing, and adapting to the landing needs of drones in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 An application environment of a method for fixed-point landing of a power patrol drone provided by an embodiment of the present invention;

[0057] Figure 2 A flowchart of a method for fixed-point landing of a power patrol UAV provided by an embodiment of the present invention;

[0058] Figure 3 A schematic diagram of the structure of a fixed-point landing system for an electric patrol drone provided by an embodiment of the present invention;

[0059] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] The method for landing a power inspection drone at a fixed point provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. When the UAV receives the landing command, the entire electric power patrol UAV fixed-point landing process starts to work together and runs in an orderly manner according to the predetermined process. Among them, the UAV end collects the color image data of the target landing platform corresponding to the UAV end in real time; locates the target landing platform according to the color image data to obtain visual positioning information; performs point cloud scanning on the target landing platform through the laser radar to obtain the radar positioning information of the target landing platform; uses the Kalman filter algorithm to fuse the visual positioning information and the radar positioning information to obtain the fused positioning information; according to the fused positioning information and the current flight information of the UAV end, the current flight information of the UAV end is adaptively adjusted to obtain the flight information adjustment amount of the UAV end; wherein, the current flight information of the UAV end includes the current flight positioning information and the current flight attitude information; according to the flight information adjustment amount of the UAV end, the control instruction of the UAV end is determined, and the control instruction of the UAV end is executed to update the current flight information of the UAV end; based on the updated current flight information of the UAV end, the step of determining the flight information adjustment amount of the UAV end based on the PID control algorithm is re-executed according to the fused positioning information and the current flight information of the UAV end, until the UAV end lands on the target landing platform.

[0062] Among them, the target landing platform can be designed as a circular structure with a diameter of 2-3 meters in terms of structural design. The main body is made of lightweight and high-strength aluminum alloy, which reduces the dead weight while ensuring the structural strength, making it easy to install and deploy. The surface is covered with a special coating with a thickness of 2 mm. The coating is made of a polymer composite material doped with nano-level rare earths (lanthanum, cerium, etc., with a doping amount between 5% and 10%). This coating can significantly enhance the reflection effect of the laser radar (it can increase the intensity of reflected light by 30%-50%), and has high diffuse reflection characteristics visually (the diffuse reflection coefficient is between 0.8-0.95, showing bright white), which is conducive to its identification by visual perception function. A circle of 3 cm high guardrails is set at the edge of the target landing platform. The guardrails are made of rubber wrapped in a metal frame. The rubber plays a buffering role, and the metal frame ensures the structural strength to prevent the drone from sliding out of the target landing platform when landing and avoid damage to the drone.

[0063] In terms of the identification setting of the target landing platform, for example, a unique identification pattern is drawn on the surface of the target landing platform in the form of a combination of concentric circles and crosshairs. The radii of the concentric circles are 0.5 meters and 1 meter respectively, and the width of the circle lines is 10 centimeters. They are drawn with fluorescent materials and can be clearly seen under different lighting conditions; the crosshairs are located in the center of the circle, with a line width of 20 centimeters and a bright red color, forming a strong visual contrast with the background, which is convenient for the visual perception function to extract features and locate and identify.

[0064] like Figure 2 As shown, the embodiment of the present application provides a method for fixed-point landing of a power patrol drone, which is applied to a drone end and includes the following steps S1 to S7. Among them:

[0065] Step S1: real-time acquisition of color image data of the target landing platform corresponding to the UAV terminal.

[0066] Among them, illustratively, the industrial-grade color camera at the bottom of the drone is started at a high frame rate (60fps) to capture real-time images of the scene below the drone, and obtain color image data including the target landing platform and the surrounding environment.

[0067] Step S2: Position the target landing platform according to the color image data to obtain visual positioning information.

[0068] The target landing platform is identified in the color image data and positioned to obtain visual positioning information of the target landing platform.

[0069] Specifically, the target landing platform is positioned according to the color image data in step S2 to obtain visual positioning information, including:

[0070] Step S201, identifying the bounding box of the target landing platform in the preprocessed color image data based on the image recognition neural network; wherein the image recognition neural network is obtained by pre-training a historical training sample set based on a deep convolutional neural network, and the historical training sample set includes historical image samples of the target landing platform and bounding boxes corresponding to the historical image samples.

[0071] Among them, a deep convolutional neural network is used to conduct massive training on the historical training sample set. During the training process, the historical image samples of the target landing platform are used as input, and the bounding boxes corresponding to the historical image samples are used as output for training.

[0072] Step S202: extracting bounding box features of the bounding box, where the bounding box features include boundary corner points and a center point.

[0073] Step S203: based on the camera intrinsic parameter matrix, convert the coordinates of the bounding box feature in the two-dimensional coordinate system into three-dimensional coordinates in the camera coordinate system.

[0074] Among them, combined with the camera intrinsic parameter matrix According to the following formula, the coordinates of the bounding box feature in the two-dimensional coordinate system are converted to the three-dimensional coordinates in the camera coordinate system to obtain:

[0075]

[0076] Where X, Y, and Z are the three-dimensional coordinates in the camera coordinate system. is the scale factor, which is obtained by solving multiple sets of known world coordinates and corresponding image coordinate points. is the coordinate of the pixel point on the plane.

[0077] Step S204: based on the camera extrinsic matrix, convert the three-dimensional coordinates of the bounding box features in the camera coordinate system into the world coordinate system, and determine the visual positioning information according to the three-dimensional coordinates of the bounding box features in the world coordinate system.

[0078] Among them, the camera external parameter matrix includes the rotation matrix from the camera coordinate system to the world coordinate system and translation vectors . Convert the three-dimensional coordinates of the bounding box feature in the camera coordinate system to the world coordinate system to obtain:

[0079]

[0080] In the formula, , , It is the three-dimensional coordinate in the world coordinate system, and the attitude angle of each feature point of the target landing platform is decomposed by the rotation matrix R. The three-dimensional coordinate and attitude angle in the world coordinate system constitute the visual positioning information.

[0081] Step S3: Scan the target landing platform by laser radar to obtain radar positioning information of the target landing platform.

[0082] Among them, the laser radar can use a high-frequency (20Hz), high-precision (±5cm) solid-state laser radar, which can be installed at the bottom of the drone to perform all-round detection of the space below in a spiral scanning mode to obtain three-dimensional point cloud data.

[0083] Among them, the target landing platform is scanned by laser radar to obtain the radar positioning information of the target landing platform, including:

[0084] Step S301: perform radar scanning on the target landing platform through a laser radar to obtain three-dimensional point cloud data of the target landing platform.

[0085] Step S302: Perform noise filtering on the three-dimensional point cloud data using a statistical filtering algorithm.

[0086] Among them, according to the statistical distribution characteristics of point cloud data, a reasonable mean is set With standard deviation The threshold (set according to the statistical analysis results of multiple pre-scan data) removes outlier noise points that obviously deviate from the main distribution of the data to ensure the reliability of the data.

[0087] Step S303: downsampling the three-dimensional point cloud data after the noise filtering process using a voxel filtering algorithm.

[0088] Among them, the voxel filtering algorithm is used for downsampling processing, and the voxel size is set (such as 0.1m×0.1m×0.1m), and the point cloud within a certain spatial range is merged into a representative point, which effectively reduces the data volume while retaining the key features of the landing platform.

[0089] Step S304: cluster the downsampled three-dimensional point cloud data based on a region growing algorithm to extract point cloud region data of the target landing platform.

[0090] Among them, based on the region growing algorithm, the point spacing threshold is set , Normal vector angle threshold , Cluster analysis is performed on the previously processed point cloud data to extract the plane point cloud area of ​​the landing platform.

[0091] Step S305: determine the normal equation according to the point cloud area data of the target landing platform, solve the normal equation using the least squares method, and obtain the normal vector and center point coordinates of the target landing platform.

[0092] The point cloud data obtained by the LiDAR contains a large amount of 3D point information of the surrounding environment, and our goal is to find the plane where the landing platform is located. Mathematically, a plane can be expressed by the equation To express.

[0093] For a given point cloud data , Ideally, these points belonging to the landing platform plane should exactly satisfy the plane equation. However, due to factors such as measurement errors, these points may not be completely on the theoretical plane.

[0094] In order to find the plane that best fits these points, the plane point cloud data is assumed to be , , so that all points are on the plane The goal is to minimize the sum of the squares of the distances and construct the objective function .

[0095] By optimizing the objective function, a set of parameters is obtained. , making the plane as close as possible to the actual landing platform plane.

[0096] Solving the above objective function to obtain the parameters Determine the normal vector of the plane where the landing platform is located The normal vector of a plane is perpendicular to the plane. For the plane equation , whose normal vector is given by The vector.

[0097] The determination of the center point coordinates is relatively complicated and is generally done in the following ways:

[0098] First, we have the plane equation .

[0099] Assuming we have a set of point cloud data belonging to the plane, we can approximate the center point of the plane by calculating the centroid (center of mass) of these point clouds. , , its centroid coordinates The calculation formula is

[0100] ,

[0102] ,

[0104]

[0105] The coordinates of the center of gravity are an approximation of the coordinates of the center point of the landing platform plane.

[0106] Therefore, the solution result Represents the normal vector, calculated by the point cloud centroid Approximately represent the center point coordinates by constructing the normal equation:

[0107]

[0108] in, , solving this equation gives the plane equation The parameters of the landing platform are used to determine the normal vector and center point coordinates of the landing platform, thereby obtaining the position and attitude information of the landing platform relative to the drone.

[0109] Among them, by constructing the normal equation, the plane can be made to conform to the overall distribution trend of the data as much as possible, reducing the influence of error points on plane determination and obtaining more accurate plane parameters.

[0110] Step S306: Determine the radar positioning information of the target landing platform relative to the UAV end according to the normal vector and the center point coordinates of the target landing platform, where the radar positioning information includes position information and attitude information.

[0111] Among them, the coordinate system is established with the drone itself as the coordinate origin, and the coordinates of the center point of the landing platform are known , by calculating the difference between this point and the origin of the drone coordinates on each coordinate axis , the translation distance of the landing platform relative to the UAV in three-dimensional space can be obtained, thereby determining its relative position.

[0112] Using the distance formula between two points in space Calculate the straight-line distance between the drone and the center point of the landing platform. The positive or negative value of determines the direction of the landing platform in the drone coordinate system.

[0113] Landing platform normal vector The angle between the coordinate system axis and the position of the platform can be used to determine the attitude angle of the platform. The angle with the z-axis is ,but

[0114]

[0115] in, is the pitch angle. Projection of the plane and The axis angle is , , is the roll angle.

[0116] Yaw angle determination: Usually the yaw angle needs to be determined by combining the flight direction of the drone and the projection of the platform normal vector on the horizontal plane. Assume that the flight direction of the drone is The positive direction of the axis, the platform normal vector is in Plane projection and The axis angle is , yaw angle According to the roll angle and related geometric relationships are calculated.

[0117] Step S4: Use the Kalman filter algorithm to fuse the visual positioning information and the radar positioning information to obtain fused positioning information.

[0118] Among them, the embodiment of the present application simultaneously receives visual positioning information and radar positioning information, and uses the Kalman filter algorithm to fuse the visual positioning information and the radar positioning information.

[0119] In step 4, the Kalman filter algorithm is used to fuse the visual positioning information and the radar positioning information to obtain the fused positioning information, including:

[0120] Step 401: construct a visual observation matrix and a radar observation matrix according to the visual positioning information and the radar positioning information.

[0121] Among them, let the visual positioning information be , the radar positioning information is , the visual observation matrix and radar observation matrix are constructed through the visual positioning information and radar positioning information.

[0122] Step 402: construct a visual observation equation based on the visual observation matrix and the system state vector.

[0123] Among them, the visual observation equation is:

[0124]

[0125] In the formula, is the visual observation value, is the visual observation matrix, is the visual observation noise, which follows a Gaussian distribution , is the visual observation noise covariance matrix.

[0126] Step 403: construct a radar observation equation according to the radar observation matrix and the system state vector;

[0127] Among them, the radar observation equation is:

[0128]

[0129] In the formula, is the radar observation value, is the lidar observation matrix, is the laser radar observation noise, which obeys Gaussian distribution , is the laser radar observation noise covariance matrix, where the system state vector is:

[0130]

[0131] In the formula, is the linear speed, is the angular velocity.

[0132] Step 404: Use the state equation to predict the state and covariance matrix at the current moment.

[0133] The state equation is:

[0134]

[0135] In the formula, is the state transfer matrix, and its value is set according to the physical characteristics and motion laws of the system; is the noise driving matrix; is the system noise, which follows Gaussian distribution , is the system noise covariance matrix.

[0136] Among them, according to the state estimation value at the previous moment To predict the current state:

[0137]

[0138] In the formula, is the estimated value of the state at the current moment.

[0139] At the same time, the prediction covariance matrix is ​​calculated as:

[0140]

[0141] In the formula, is the covariance matrix at the current moment, and T is the matrix transpose.

[0142] Step 405: Update the Kalman gain through the visual observation equation and the radar observation equation.

[0143] Among them, the updated Kalman gain is:

[0144]

[0145] In the formula, is the observation matrix, It is determined based on a combination of visual and lidar observations. is the observation noise covariance matrix, is the visual observation noise covariance matrix and the lidar observation noise covariance matrix It is synthesized based on factors such as information fusion weight.

[0146] Step 406: Use the Kalman gain to update the current state and covariance matrix until the update iteration converges, and output the state after the update iteration converges as the fused positioning information.

[0147] Among them, the state and covariance matrix of the current moment are updated respectively:

[0148]

[0149]

[0150] After multiple rounds of iterative optimization, the fused positioning information is output. The fused positioning information includes position information and posture information, which is recorded as .

[0151] Step S5: Based on the PID control algorithm, the current flight information of the drone is adaptively adjusted according to the fused positioning information and the current flight information of the drone, so as to obtain the flight information adjustment amount of the drone; wherein the current flight information of the drone includes the current flight positioning information and the current flight attitude information.

[0152] The drone's rotor motors and servos are driven based on the fused positioning information to achieve precise control of the drone's flight attitude and position. The adaptive proportional-integral-differential control algorithm is used, which can dynamically adjust the control parameters according to the drone's real-time flight status and environmental conditions to ensure the stability and accuracy of the drone during landing.

[0153] Among them, the fused positioning information includes the location information of the target landing platform; the flight information adjustment amount includes the position adjustment amount and the angle adjustment amount.

[0154] In step S5, based on the PID control algorithm, the current flight information of the UAV is adaptively adjusted according to the fused positioning information and the current flight information of the UAV, and the flight information adjustment amount of the UAV is obtained, including:

[0155] Step S501: determining the position deviation between the current position of the drone and the position information of the target landing platform.

[0156] Among them, the position deviation can be divided into horizontal position deviation and vertical position deviation.

[0157] Step S502: Adaptively adjust the current position of the drone end through a PID control algorithm according to the position deviation to obtain a position adjustment amount; wherein the position adjustment amount includes a horizontal position adjustment amount and a vertical position adjustment amount.

[0158] In terms of horizontal position control, the target position coordinates of the landing platform are , real-time monitoring of the deviation between the current position of the drone and the target position , and the rate of change of the deviation According to the preset scale factor , integral coefficient , differential coefficient ,

[0159] According to the formula The speed adjustment of each rotor motor is accurately calculated to drive the drone to approach the landing platform smoothly in the horizontal direction.

[0160] In terms of vertical position control, the target height of the landing platform is , real-time monitoring of the deviation between the current altitude of the drone and the target altitude and vertical velocity deviation , set the proportionality factor , integral coefficient , differential coefficient , according to the formula The adjustment amount of the total lift is calculated to control the descent speed and position of the UAV in the vertical direction, so that the UAV can descend smoothly above the target landing platform.

[0161] Step S503: Adaptively adjust the current angle deviation of the drone end according to the angle deviation between the current posture of the drone end and the expected posture through the PID control algorithm to obtain an angle adjustment amount; the angle adjustment amount includes the Euler angle adjustment amount, the pitch angle adjustment amount and the roll angle adjustment amount.

[0162] In terms of attitude control, the expected Euler angle is , real-time monitoring of the deviation between the current Euler angle of the drone and the expected Euler angle , , And the corresponding angular velocity deviation , set the proportionality factor , integral coefficient , differential coefficient , taking the yaw angle as an example, according to the formula Calculate the servo angle adjustment. Similarly, the pitch and roll angles can be calculated. By precisely adjusting the servo angles, the drone can be accurately aligned with the landing platform to ensure a smooth landing.

[0163] Step S6: determine the control instruction of the drone end according to the flight information adjustment amount of the drone end, execute the control instruction of the drone end, and update the current flight information of the drone end.

[0164] Step S7: Based on the updated current flight information of the drone, re-execute the PID control algorithm to determine the flight information adjustment amount of the drone according to the fused positioning information and the current flight information of the drone, until the drone lands on the target landing platform.

[0165] The flight information adjustment amount on the drone side is converted into corresponding control instructions, which are sent to the drone's electronic speed regulator and servo controller to drive the rotor motor and servo to achieve precise control of the drone's flight attitude and position. During the entire landing process, the flight control module repeats the above steps, and adjusts the control instructions in real time according to the new fused positioning information and flight status, so that the drone gradually approaches and stabilizes above the landing platform.

[0166] It should be noted that the embodiment of the present application obtains fused positioning information by collecting color image data of the target landing platform corresponding to the drone end in real time, and uses the Kalman filter algorithm to fuse the visual positioning information and the radar positioning information. Based on the PID control algorithm, the current flight information of the drone end is adaptively adjusted according to the fused positioning information and the current flight information of the drone end, and the control instructions of the drone end are determined and executed according to the adjustment amount of the flight information of the drone end, and the current flight information of the drone end is updated. The flight information is continuously updated during the landing process until the drone end lands on the target landing platform, thereby improving the accuracy and reliability of the drone's fixed-point landing, and adapting to the landing needs of drones in complex environments.

[0167] In some embodiments, the method further includes preprocessing the color image data. Preprocessing the color image data includes:

[0168] Step S211: grayscale the color image data to obtain a grayscale image.

[0169] Step S212: determine the grayscale value of each pixel in the grayscale image by weighted average method.

[0170] Among them, the grayscale value of each pixel is determined by the weighted average method, which simplifies the data processing and preliminarily outlines the contour of the landing platform.

[0171] Step S213: performing noise filtering on the grayscale image by using Gaussian filtering and median filtering according to the grayscale value of each pixel in the grayscale image.

[0172] Among them, the grayscale image is denoised by combining Gaussian filtering and median filtering. Gaussian filtering smoothes the image based on the normal distribution function to reduce the influence of Gaussian noise; median filtering replaces the current pixel value with the median of the grayscale values ​​of the neighboring pixels, effectively removing salt and pepper noise and purifying the image quality.

[0173] Step S214: Use the Sobel algorithm to enhance the edge features of the grayscale image after the noise filtering process.

[0174] Among them, the Sobel algorithm is introduced to enhance the edge features, at the pixel point At, according to the convolution kernel and grayscale matrix in the horizontal and vertical directions Perform operations to obtain the gradient values ​​in the horizontal and vertical directions, and then calculate the gradient amplitude , which greatly enhances the edge contour of the landing platform, making it more clearly visible in the image and facilitating subsequent feature extraction and positioning operations.

[0175] In some embodiments, a plurality of pressure sensors are evenly distributed on the target landing platform, and the method further includes:

[0176] Step S801: monitoring the pressure information of the drone after landing on the target landing platform according to multiple pressure sensors;

[0177] Step S802: When the pressure information monitored by the multiple pressure sensors reaches or exceeds the preset pressure threshold, it is determined that the landing operation of the drone end landing on the target landing platform is successful, and a landing success signal is generated and sent to the drone end;

[0178] Step S803: When the drone receives a landing success signal, the flight operation of the drone is turned off, so that the drone stops on the target landing platform.

[0179] For example, when the landing gear of the drone contacts the landing platform, the four high-precision pressure sensors (accuracy 0.05kg, range 0 - 50kg) evenly distributed on the bottom of the platform are sensitive to pressure changes. Once the pressure reaches the preset threshold (such as 1kg, which can be adjusted according to the actual weight of the drone), the landing success signal is immediately triggered. The signal is quickly transmitted to the drone flight control system by wireless transmission through a wireless communication module with a 2.4GHz frequency band and a communication distance of 1 km set inside the platform. After receiving the signal, the flight control system immediately shuts down and locks each rotor motor to ensure that the drone is firmly parked on the landing platform. At this point, the fixed-point landing flow is completed.

[0180] Based on the same inventive concept, an embodiment of the present application also provides a power patrol drone fixed-point landing system for implementing the above-mentioned power patrol drone fixed-point landing method.

[0181] The implementation solution to the problem provided by the system is similar to the implementation solution recorded in the above method. Therefore, the specific limitations in one or more embodiments of the power patrol drone fixed-point landing system provided below can be referred to the limitations on the power patrol drone fixed-point landing method above, and will not be repeated here.

[0182] like Figure 3 As shown, the embodiment of the present application also provides a power patrol drone fixed-point landing system, which is applied to the drone end, and the drone end is provided with a visual perception module 100, a visual positioning module 200, a laser radar ranging module 300, a data fusion module 400 and a flight control module 500;

[0183] The visual perception module 100 is used to collect color image data of the target landing platform corresponding to the UAV end in real time;

[0184] The visual positioning module 200 is used to locate the target landing platform according to the color image data to obtain visual positioning information;

[0185] The laser radar ranging module 300 is used to perform point cloud scanning on the target landing platform through the laser radar to obtain radar positioning information of the target landing platform;

[0186] The data fusion module 400 is used to fuse the visual positioning information and the radar positioning information using a Kalman filter algorithm to obtain fused positioning information;

[0187] The flight control module 500 is used to adaptively adjust the current flight information of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end, so as to obtain the flight information adjustment amount of the drone end; wherein the current flight information of the drone end includes the current flight positioning information and the current flight attitude information; it is also used to determine the control instructions of the drone end according to the flight information adjustment amount of the drone end, and execute the control instructions of the drone end to update the current flight information of the drone end; it is also used to re-execute the process of determining the flight information adjustment amount of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end based on the updated current flight information of the drone end, until the drone end lands on the target landing platform.

[0188] like Figure 4 As shown, an embodiment of the present application also provides an electronic device, the electronic device 10 includes a memory 20 and a processor 30, the memory 20 stores a computer program, and when the computer program is executed by the processor 30, the processor 30 executes the steps of the fixed-point landing method of the power patrol drone in any of the above embodiments.

[0189] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the steps of the method for fixed-point landing of a power inspection drone as in any of the above embodiments are implemented.

[0190] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, electronic device and computer storage medium can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0191] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatus.

[0192] In several embodiments provided by the present invention, it is understood that each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and a part of a module, a program segment or a code includes one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.

[0193] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, electronic devices, computer storage media and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0194] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0195] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0196] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for executing all or part of the steps of the method described in each embodiment of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk, etc. Various media that can store program codes.

[0197] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fixed-point landing of a power inspection UAV, applied to the UAV end, characterized in that: include: Collecting color image data of the target landing platform corresponding to the UAV end in real time; Positioning the target landing platform according to the color image data to obtain visual positioning information; Performing point cloud scanning on the target landing platform by using a laser radar to obtain radar positioning information of the target landing platform; Using a Kalman filter algorithm to fuse the visual positioning information and the radar positioning information to obtain fused positioning information; Based on the PID control algorithm, the current flight information of the drone end is adaptively adjusted according to the fused positioning information and the current flight information of the drone end to obtain the flight information adjustment amount of the drone end; wherein the current flight information of the drone end includes current flight positioning information and current flight attitude information; Determine the control instruction of the drone end according to the flight information adjustment amount of the drone end, execute the control instruction of the drone end, and update the current flight information of the drone end; Based on the updated current flight information of the drone end, the PID control algorithm is re-executed to determine the step of adjusting the flight information of the drone end according to the fused positioning information and the current flight information of the drone end until the drone end lands on the target landing platform.

2. The method for fixed-point landing of a power inspection drone according to claim 1 is characterized in that: It also includes preprocessing the color image data; The preprocessing of the color image data comprises: Performing grayscale processing on the color image data to obtain a grayscale image; Determine the grayscale value of each pixel in the grayscale image by weighted average method; Performing noise filtering on the grayscale image by Gaussian filtering and median filtering according to the grayscale value of each pixel in the grayscale image; The Sobel algorithm is used to enhance the edge features of the grayscale image after noise filtering.

3. The method for fixed-point landing of a power inspection UAV according to claim 1 is characterized in that: Positioning the target landing platform according to the color image data to obtain visual positioning information includes: Identify a bounding box of the target landing platform in the preprocessed color image data based on an image recognition neural network; wherein the image recognition neural network is obtained by pre-training a historical training sample set based on a deep convolutional neural network, and the historical training sample set includes historical image samples of the target landing platform and bounding boxes corresponding to the historical image samples; Extracting bounding box features of the bounding box, wherein the bounding box features include bounding corner points and a center point; Based on the camera intrinsic parameter matrix, the coordinates of the bounding box feature in the two-dimensional coordinate system are converted into three-dimensional coordinates in the camera coordinate system; Based on the camera extrinsic matrix, the three-dimensional coordinates of the bounding box features in the camera coordinate system are converted to the world coordinate system, and the visual positioning information is determined according to the three-dimensional coordinates of the bounding box features in the world coordinate system.

4. The method for fixed-point landing of a power inspection UAV according to claim 1 is characterized in that: The step of performing point cloud scanning on the target landing platform by using a laser radar to obtain radar positioning information of the target landing platform includes: Scanning the target landing platform by a laser radar to obtain three-dimensional point cloud data of the target landing platform; Performing noise filtering on the three-dimensional point cloud data using a statistical filtering algorithm; The voxel filtering algorithm is used to downsample the three-dimensional point cloud data after noise filtering; Performing clustering processing on the downsampled three-dimensional point cloud data based on a region growing algorithm to extract point cloud region data of the target landing platform; Determine a normal equation according to the point cloud area data of the target landing platform, and solve the normal equation using the least square method to obtain the normal vector and center point coordinates of the target landing platform; The radar positioning information of the target landing platform relative to the UAV end is determined according to the normal vector and the center point coordinates of the target landing platform, and the radar positioning information includes position information and attitude information.

5. The method for fixed-point landing of a power inspection UAV according to claim 1 is characterized in that: The method of fusing the visual positioning information and the radar positioning information using a Kalman filter algorithm to obtain fused positioning information includes: Constructing a visual observation matrix and a radar observation matrix according to the visual positioning information and the radar positioning information; Constructing a visual observation equation according to the visual observation matrix and the system state vector; Constructing a radar observation equation according to the radar observation matrix and the system state vector; Use the state equation to predict the current state and covariance matrix; Update the Kalman gain by using the visual observation equation and the radar observation equation; The Kalman gain is used to update the state and covariance matrix at the current moment until the update iteration converges, and the state after the update iteration converges is output as the fused positioning information.

6. The method for fixed-point landing of a power inspection drone according to claim 1 is characterized in that: Based on the PID control algorithm, the fused positioning information includes the position information of the target landing platform; the flight information adjustment amount includes the position adjustment amount and the angle adjustment amount; The method based on the PID control algorithm, adaptively adjusting the current flight information of the drone end according to the fused positioning information and the current flight information of the drone end, and obtaining the flight information adjustment amount of the drone end, includes: According to the position deviation between the current position of the drone and the position information of the target landing platform; Adaptively adjusting the current position of the drone end through a PID control algorithm according to the position deviation to obtain a position adjustment amount; wherein the position adjustment amount includes a horizontal position adjustment amount and a vertical position adjustment amount; According to the angle deviation between the current posture of the drone and the expected posture, the current angle deviation of the drone is adaptively adjusted through a PID control algorithm to obtain an angle adjustment amount; the angle adjustment amount includes an Euler angle adjustment amount, a pitch angle adjustment amount, and a roll angle adjustment amount.

7. The method for fixed-point landing of a power inspection drone according to claim 1 is characterized in that: A plurality of pressure sensors are evenly distributed on the target landing platform, and the method further comprises: Monitoring the pressure information of the drone after landing on the target landing platform according to the plurality of pressure sensors; When the pressure information monitored by the plurality of pressure sensors reaches or exceeds a preset pressure threshold, it is determined that the landing operation of the drone end to land on the target landing platform is successful, and a landing success signal is generated and sent to the drone end; When the drone receives the landing success signal, the flight operation of the drone is turned off, so that the drone stops on the target landing platform.

8. A fixed-point landing system for electric power inspection drones, applied to drone terminals, characterized in that: The drone end is provided with a visual perception module, a visual positioning module, a laser radar ranging module, a data fusion module and a flight control module; The visual perception module is used to collect color image data of the target landing platform corresponding to the drone end in real time; The visual positioning module is used to locate the target landing platform according to the color image data to obtain visual positioning information; The laser radar ranging module is used to perform point cloud scanning on the target landing platform through the laser radar to obtain radar positioning information of the target landing platform; The data fusion module is used to fuse the visual positioning information and the radar positioning information using a Kalman filter algorithm to obtain fused positioning information; The flight control module is used to adaptively adjust the current flight information of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end, so as to obtain the flight information adjustment amount of the drone end; wherein the current flight information of the drone end includes the current flight positioning information and the current flight attitude information; it is also used to determine the control instructions of the drone end according to the flight information adjustment amount of the drone end, and execute the control instructions of the drone end to update the current flight information of the drone end; it is also used to re-execute the process of determining the flight information adjustment amount of the drone end based on the PID control algorithm according to the fused positioning information and the current flight information of the drone end based on the updated current flight information of the drone end, until the drone end lands on the target landing platform.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the processor executes the steps of the method for fixed-point landing of a power patrol drone as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the steps of the method for fixed-point landing of a power inspection drone as described in any one of claims 1 to 7 are implemented.

Images