A fixed-wing aircraft glide control method based on pixel target feedback

Abstract

The application belongs to the field of unmanned aerial vehicle flight control, and particularly relates to a fixed-wing aircraft glide control method based on pixel target feedback. The method extracts the pixel coordinates of the runway from the image, and corrects the pixel coordinates by combining the unmanned aerial vehicle attitude through a correction algorithm. The corrected pixel coordinates can approximately describe the relative position of the given glide path and the aircraft. The airspeed command and the track angle command are used as the control output, thereby solving the aircraft glide path keeping problem under the condition that the visual guidance cannot obtain obvious image features when far away from the airport runway, and effectively expanding the distance of the visual guidance.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) flight control, specifically relating to a fixed-wing aircraft glide control method based on pixel target feedback. Background Technology

[0002] Current visual guidance technology for drone landing is limited by the imaging quality of the image features captured by the visual sensors. The effective distance from the runway to obtain accurate positioning information through visual recognition at the end of the landing process is about 1km, which is far from sufficient for the full-process visual landing of large drones. Summary of the Invention

[0003] Purpose of the invention: To provide a fixed-wing aircraft glide control method based on pixel target feedback, so as to enable the UAV to glide autonomously and complete the landing under limited visual guidance information.

[0004] Technical solution:

[0005] A fixed-wing aircraft glide control method based on pixel target feedback includes:

[0006] Step 1: When the distance between the drone and the designated landing point is within a preset range, obtain an image containing the airport runway from the video information of the drone's forward-facing camera;

[0007] Step 2: Extract the approximate features of the airport runway from the image in Step 1, and fit the extracted runway using a single straight line segment to determine the coordinates A0(u, v) of the endpoint on the side closer to the landing direction in the pixel plane.

[0008] Step 3: Use the UAV attitude angle information and the parameters of the UAV's forward-looking camera to correct the endpoint coordinates A0(u, v) on the side of the runway closest to the landing direction obtained in Step 2, and obtain the corrected endpoint coordinates A(u1, v1).

[0009] Step 4: Calculate the difference between the corrected pixel x-coordinate u1 of the runway endpoint A and the x-coordinate u0 of the camera optical center, and add the lateral telemetry amount u. b We get u1+u b -u0, multiplied by the comprehensive conversion gain k1, combined with the landing airport runway heading, and after a limiting process, the UAV track angle command ψ is obtained. cmd ;

[0010] Step 5: Based on the given ideal glide angle γ e Calculate the vertical coordinate v2 of the landing point N of the UAV at the current position with an ideal glide angle in the camera pixel plane, and calculate the difference between its vertical coordinate and the runway endpoint A, plus the longitudinal telemetry amount v. b v1+v b-v2, and multiplied by the overall conversion gain k2, based on the ideal glide angle γ e The ideal azimuth speed is calculated from the total speed V of the UAV, and the azimuth speed command Vz of the UAV is obtained by combining the results. cmd ;

[0011] Step 6: Send the upward speed command Vz cmd and track angle command ψ cmd As a control output, it guides the drone to complete the entire descent and landing process.

[0012] Furthermore, in step 1, the preset range is 4-8 kilometers.

[0013] Furthermore, in step 1, the airport runway image is a single-frame image in RGB format.

[0014] Furthermore, step 3 specifically includes:

[0015] Step 3-1: Collect the current pitch angle, roll angle, and yaw angle of the UAV;

[0016] Step 3-2: Obtain the intrinsic parameter matrix K of the UAV's forward-looking camera;

[0017] Step 3-3: Use a virtual gimbal method to correct the coordinates A0 of the runway endpoint M in the pixel plane. Based on the fact that the coordinates of the runway endpoint M and the UAV in the world coordinate system are unique and invariant at the same moment, the equation can be established: W′=KR′R T K -1 W, where W is the extended dimension vector of A0(u, v), has K is the intrinsic parameter matrix of the airborne forward-looking camera; R is the attitude matrix corresponding to the current attitude of the UAV; R′ is the attitude matrix corresponding to the virtual attitude of the UAV; W′ is the coordinate of the runway endpoint M in the pixel plane under the virtual attitude. That is, the corresponding corrected coordinate point A(u1, v1).

[0018] Furthermore, in step 4: k1 = k 1Y ·k 1V Wherein, the conversion gain k 1Y It converts in-plane pixel deviation into spatial lateral position deviation;

[0019] Conversion gain k 1V It converts the lateral positional deviation in space into an upward velocity deviation based on the performance of the UAV.

[0020] Furthermore, in step 4, k 1Y The size is inversely proportional to the distance between the drone and the landing point, which is qualitatively described by runway imaging line segments or collected by other sensors.

[0021] Furthermore, in step 4, k 1V The gain value is determined by the drone's performance, and the higher the drone's speed, the smaller the gain value.

[0022] Furthermore, in step 5, based on the given ideal glide angle γ... e Calculate the vertical coordinate v2 of the landing point N of the UAV at the current position with an ideal glide angle in the camera pixel plane, specifically including:

[0023] Obtain the ideal glide angle γ of the drone e Camera vertical resolution v plx The field of view τ perpendicular to the camera;

[0024] The ideal glide angle γ of the drone e Camera vertical resolution v pix Substitute the field of view τ perpendicular to the camera's vertical direction into the formula. Calculate the ideal glide angle γ for the drone's current position. e The landing point N of the glide path has a vertical coordinate v2 in the camera pixel plane.

[0025] Beneficial effects:

[0026] This application extracts the pixel coordinates of the runway from the image and corrects them by combining the UAV attitude with a correction algorithm. The corrected pixel coordinates can approximately describe the relative position of a given glide path and the aircraft. The upward speed command and the track angle command are used as control outputs, thereby solving the problem of aircraft glide path maintenance when visual guidance cannot obtain obvious image features at a distance from the airport runway, and effectively extending the application distance of visual guidance. Attached Figure Description

[0027] Figure 1 The lateral track angle command ψ cmd ;

[0028] Figure 2 The vertical upward velocity command Vz cmd ;

[0029] Figure 3 This is a schematic diagram illustrating the physical meaning of v2. Detailed Implementation

[0030] While visual guidance may struggle to precisely extract runway edge contours and establish accurate relative positioning coordinates at a distance of 1 km, it can still generally visualize the runway's approximate direction and imaging plane coordinates. This allows for control of the aircraft's trajectory, extending the maximum effective range of visual guidance to 4-8 km. This invention utilizes visual information with limited accuracy to calculate UAV trajectory commands for glide landing control, enabling autonomous glide and landing of the UAV with limited visual guidance information.

[0031] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. This application provides a fixed-wing aircraft glide control method based on pixel target feedback, the method comprising:

[0032] Step 1: Taking the drone's glide landing process as the initial situation, when the drone is within the preset range from the designated landing point, activate the visual guidance device to obtain image information including the airport runway from the video information of the drone's forward-facing camera;

[0033] Specifically, depending on the camera's focal length, the preset range is 4-8 kilometers.

[0034] In practical applications, airport runway image information is a single-frame image in RGB format.

[0035] Step 2: Extract the approximate features of the airport runway from the image in Step 1, and fit the extracted runway using a single straight line segment to determine the coordinates A0(u, v) of the endpoint on the side closer to the landing direction in the pixel plane;

[0036] Specifically, deep learning-based image recognition and segmentation algorithms are trained using actual flight videos or simulation scenarios modeled closely to reality, thereby obtaining a recognition and segmentation neural network with high generalization ability. The imaging line segments corresponding to the airport runway are extracted and fitted, and the coordinates A0(u, v) of the runway endpoint on the side closer to the landing point are determined.

[0037] Step 3: Use the extracted UAV attitude angle information and UAV onboard camera parameters to correct the endpoint coordinates A0(u, v) on the side of the runway closest to the landing direction obtained in Step 2, and obtain the corrected endpoint coordinates A(u1, v1).

[0038] Step 3 specifically includes:

[0039] Step 3-1: Collect the current pitch angle, roll angle, and yaw angle of the UAV;

[0040] In practical applications, the attitude angle of the drone can be obtained through sensors such as inertial navigation. In harsh environments, features such as the horizon can be extracted from images captured by cameras for rough calculations.

[0041] Step 3-2: Obtain the intrinsic parameter matrix K of the UAV's forward-looking camera;

[0042] Step 3-3: Use a virtual gimbal method to correct the coordinates A0 of the runway endpoint M in the pixel plane. Since the coordinates of the runway endpoint M and the UAV in the world coordinate system are unique and invariant at the same moment, an equation can be established:

[0043] W′=KR′RT K -1 W

[0044] Where W is the extended dimension vector of A0(u, v), we have

[0045] K is the intrinsic parameter matrix of the airborne forward-looking camera;

[0046] R is the attitude matrix corresponding to the current attitude of the UAV;

[0047] R′ is the attitude matrix corresponding to the virtual attitude of the UAV;

[0048] W′ represents the coordinates of the runway endpoint M in the pixel plane under the virtual attitude. That is, the corresponding corrected coordinate point A(u1, v1).

[0049] Step 4: Calculate the difference between the corrected horizontal coordinate u1 of the runway endpoint A pixel and the horizontal coordinate u0 of the camera optical center, and add the lateral telemetry amount u. b We get u1+u b -u0, multiplied by the comprehensive conversion gain k1, and combined with the landing airport runway heading, after passing through steps such as amplitude limiting, the UAV track angle command ψ is obtained. cmd ;

[0050] Combination Figure 1 As shown, step 4 specifically includes:

[0051] Step 4-1: Obtain the abscissa u0 of the camera's optical center and the heading ψ0 of the landing airport runway;

[0052] Step 4-2: Determine the lateral telemetry amount u b Adjusting this value can change the landing sideslip of the drone;

[0053] In practical applications, the lateral telemetry amount u b This information was provided by the pilot after observation at the ground station.

[0054] Step 4-3: Calculate the UAV track angle command ψ using the above variables according to the block diagram. cmd ;

[0055] The overall conversion gain k1 acts on u1+u b -u0 term, k1=k 1Y ·k 1V ;

[0056] Conversion gain k 1Y The physical meaning is to convert in-plane pixel deviation into spatial lateral position deviation. The magnitude of this gain is inversely proportional to the distance between the UAV and the landing point. This distance is qualitatively described by the runway imaging line segment or acquired by other sensors.

[0057] Conversion gain k 1V The physical meaning is to convert the lateral position deviation in space into the axial velocity deviation based on the performance of the UAV. This gain value is determined by the performance of the UAV, and the higher the speed of the UAV, the smaller the gain value.

[0058] Step 5: Based on the given ideal glide angle γ e Calculate the vertical coordinate v2 of the landing point N of the UAV at the current position with an ideal glide angle in the camera pixel plane, and calculate the difference between its vertical coordinate and the runway endpoint A, plus the longitudinal telemetry amount v. b v1+v b -v2, and multiplied by the overall conversion gain k2, based on the ideal glide angle γ e The ideal azimuth speed is calculated from the total speed V of the UAV, and the azimuth speed command Vz of the UAV is obtained by combining the results. cmd .

[0059] Combination Figure 2 As shown, step 5 specifically includes:

[0060] Step 5-1: Obtain the ideal glide angle γ for the UAV e Camera vertical resolution v pix The field of view τ perpendicular to the camera;

[0061] Step 5-2: Set the ideal glide angle γ of the UAV e Camera vertical resolution v pix Substitute the field of view τ perpendicular to the camera's vertical direction into the formula. Calculate the ideal glide angle γ for the drone's current position. e The landing point N of the glide is located at the vertical coordinate v2 in the camera pixel plane;

[0062] Step 5-3: Determine the longitudinal telemetry amount v b The pilot adjusts the longitudinal telemetry value v at the ground station. b This changes the drone's landing point, which is represented as v1+v in the pixel plane. b ;

[0063] Step 5-4: Collect the total speed V of the drone;

[0064] Step 5-5: Calculate the above variables according to the block diagram to obtain the upward velocity command Vz. cmd

[0065] Furthermore, the overall conversion gain k2 acts on v1+v b -v2 term, k2 = k 2H ·k 2V ;

[0066] Furthermore, the conversion gain k 2HThe physical meaning is to convert in-plane pixel deviation into spatial height deviation. The magnitude of this gain is inversely proportional to the distance between the UAV and the landing point. This distance is qualitatively described by the runway imaging line segment or acquired by other sensors.

[0067] Furthermore, the conversion gain k 2V The physical meaning is to convert the spatial altitude deviation into a flight path angle deviation based on the performance of the UAV, and this gain value is determined by the performance of the UAV.

[0068] Step 6: Send the celestial velocity command Vz cmd and track angle command ψ cmd As a control output, it guides the drone to complete the entire descent and landing process.

[0069] Specifically, the celestial velocity command Vz is calculated through steps 4 and 5. cmd and track angle command ψ cmd By combining the current azimuth speed and current trajectory angle of the UAV, feedback signals are generated and applied to the longitudinal control outer loop and the lateral control outer loop of the UAV, respectively, to generate control surface commands, thereby changing the position of the UAV and controlling and guiding the UAV to complete the entire glide landing process.

Claims

1. A fixed-wing aircraft glide control method based on pixel target feedback, characterized in that, include: Step 1: When the distance between the drone and the designated landing point is within a preset range, obtain an image containing the airport runway from the video information of the drone's forward-facing camera; Step 2: Extract the features of the airport runway from the image in Step 1, and fit the extracted runway using a single straight line segment to determine the coordinates of the endpoint on the side closest to the landing direction in the pixel plane. ; Step 3: Using the UAV attitude angle information and the parameters of the UAV's forward-facing camera, analyze the endpoint coordinates of the runway on the landing side obtained in Step 2. After correction, the endpoint coordinates are obtained. ; Step 4: Calculate the corrected pixel x-coordinate of runway endpoint A. x-coordinate of the camera's optical center The difference plus the lateral telemetry ,get And multiplied by the overall conversion gain Based on the landing airport runway heading, the UAV's flight path angle command is obtained after the width limiting process. ; Step 5: Based on the given ideal glide angle Calculate the vertical coordinate of the landing point N of the UAV at its current position with an ideal glide angle in the camera pixel plane. Calculate the difference between its longitudinal coordinate and the longitudinal telemetry of runway endpoint A, plus the longitudinal telemetry amount. , And multiplied by the overall conversion gain According to the ideal glide angle The ideal azimuth speed is calculated from the total speed V of the UAV, and the azimuth speed command of the UAV is obtained by combining these calculations. ; Step 6: Send the celestial speed command and track angle command As a control output, it guides the drone to complete the entire descent and landing process.

2. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, In step 1, the preset range is 4-8 kilometers.

3. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, In step 1, the airport runway image is a single frame image in RGB format.

4. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, Step 3 specifically includes: Step 3-1: Collect the current pitch angle, roll angle, and yaw angle of the UAV; Step 3-2: Obtain the intrinsic parameter matrix K of the UAV's forward-looking camera; Step 3-3: Use a virtual gimbal method to determine the coordinates of the runway endpoint M in the pixel plane. Make corrections based on the runway endpoints at the same time. The coordinates of the drone in the world coordinate system are unique and invariant, and an equation can be established: ,in, for Extended dimension vector, has ; This is the intrinsic parameter matrix of the airborne forward-looking camera; This is the attitude matrix corresponding to the current attitude of the UAV; The attitude matrix corresponding to the virtual attitude of the UAV; Let M be the coordinates of the runway endpoint M in the pixel plane under the virtual attitude. That is, the corresponding corrected coordinate points .

5. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, In step 4: Among them, conversion gain It converts in-plane pixel deviation into spatial lateral position deviation; Conversion gain It converts the lateral positional deviation in space into an upward velocity deviation based on the performance of the UAV.

6. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 5, characterized in that, In step 4, The size is inversely proportional to the distance between the drone and the landing point, which is qualitatively described by runway imaging line segments or collected by other sensors.

7. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, In step 4, The gain value is determined by the performance of the drone, and the higher the speed of the drone, the smaller the gain value.

8. The fixed-wing aircraft glide control method based on pixel target feedback according to claim 1, characterized in that, In step 5, based on the given ideal glide angle... Calculate the vertical coordinate of the landing point N of the UAV at its current position with an ideal glide angle in the camera pixel plane. Specifically, it includes: Achieving the ideal glide angle for the drone Camera vertical resolution Field of view in the vertical direction of the camera ; The ideal glide angle of the drone Camera vertical resolution Field of view in the vertical direction of the camera Substitute into the formula Calculate the drone's current position to maintain the ideal glide angle The vertical coordinate of the landing point N in the camera pixel plane is... .

Images