Rotor disc inclination angle measurement method based on binocular vision

Abstract

The invention belongs to the technical field of helicopter model rotor tests, and particularly relates to a binocular vision-based rotor disc inclination angle measurement method. The method comprises the following steps: S1, erecting a binocular stereoscopic vision camera and an LED light source above a model rotor test bed, wherein the camera vertically and downwards shoots reflective targets adhered to a rotor hub and a support arm; a photoelectric sensor is installed right opposite to a rotating shaft of the model rotor wing test bed, a camera is triggered by using a TOP signal generated by the photoelectric sensor, and the orientation of an image shot by the camera is identified; s2, identifying a reflective target pasted on the rotor hub in the image shot by the camera, and calibrating the binocular camera to obtain a calibration matrix in combination with the relative position information of the known hub mark point; s3, determining three-dimensional coordinates of all propeller hub mark points and support arm mark points; and S4, fitting the three-dimensional coordinates of the marking points of the propeller hub into a plane calculation normal vector, fitting the three-dimensional coordinates of the same marking point on the four support arms in different directions into a plane calculation normal vector respectively, and obtaining the inclination angle of the rotor disc.

Description

Technical Field

[0001] The invention belongs to the technical field of helicopter model rotor test, and in particular relates to a rotor blade disc inclination angle measurement method based on binocular vision. Background Art

[0002] In wind tunnel tests, the drag coefficient error caused by the measurement error of the model's angle of attack accounts for about a quarter of the total drag coefficient error. The main existing measurement methods are angle sensors, laser goniometers, and video measurement technology. Angle sensors and laser goniometers have wiring on rotating parts and sensor pasting, which greatly interfere with the aerodynamic shape of the model. Video measurement technology is divided into monocular measurement and binocular measurement. The monocular measurement calibration is relatively simple, but the measurement accuracy is not stable enough; the binocular measurement calibration is more complicated and the single precision is higher. In order to improve the measurement accuracy of the model rotor disc inclination angle, this paper adopts a non-contact optical measurement method, combined with the measurement mark points for self-calibration to obtain a calibration matrix, and performs real-time measurement of the disc inclination angle based on this calibration matrix. Summary of the invention

[0003] The technical problem solved by the present invention is to accurately measure the inclination angle of the model rotor blade disc.

[0004] The technical solution of the present invention:

[0005] A method for measuring the inclination angle of a rotor blade disc based on binocular vision, the method comprising:

[0006] S1, a binocular stereo vision camera and an LED light source are set up above the model rotor test bench, and the camera shoots the reflective target attached to the rotor hub and the support arm vertically downward; a photoelectric sensor is installed facing the rotating axis of the model rotor test bench, and the TOP signal generated by the photoelectric sensor is used to trigger the camera and identify the orientation of the camera shooting image;

[0007] S2, identifying the reflective target pasted on the rotor hub in the image captured by the camera, and combining the relative position information of the known hub marking points to calibrate the binocular camera to obtain a calibration matrix;

[0008] S3, determining the three-dimensional coordinates of all hub marking points and arm marking points;

[0009] S4, the three-dimensional coordinates of the hub mark point are fitted into a plane to calculate the normal vector, and the three-dimensional coordinates of the same mark point on the four arms at different orientations are respectively fitted into a plane to calculate the normal vector to obtain the rotor disc inclination angle.

[0010] Furthermore, S1 is specifically:

[0011] Two high-speed cameras and an LED light source are set up vertically downward 2m above the model rotor test bench; a photoelectric sensor is installed opposite to the rotating axis of the test bench, and the photoelectric sensor is connected to the high-speed camera and the LED light source through a frequency doubling synchronizer.

[0012] Furthermore, the focal length of the camera lens is 50 mm, and the field of view size is 1 m*0.625 m; the camera image resolution is set to 2048*1280 pixels and the exposure time is 30 microseconds, the frequency synchronization device is set to 120 times, and the camera external trigger mode is turned on to collect images.

[0013] Furthermore, before S2, the method further includes:

[0014] Adjust the camera vertically downward to ensure that the rotor hub and the reflective target attached to the arm are captured within ±15°.

[0015] Furthermore, S2 is specifically:

[0016] In the hovering state with a speed of 50 rpm, multiple images containing hub markers are collected at each rotor shaft inclination phase;

[0017] The YOLOv5 network model is used to identify and distinguish the arm markers and hub markers in the image, and the EDCircles circle detection method is used to detect the center coordinates of all markers.

[0018] Using the relative position information of the hub marking points obtained in the image and the known hub marking points, a physical template is established in the form of polar coordinates to calibrate the binocular camera. The intrinsic parameter matrices of the two cameras, the rotation matrix and the translation matrix between the two cameras are obtained as the calibration matrix.

[0019] Furthermore, S3 is specifically:

[0020] Under a certain test state, images containing hub marks and arm marks are collected, and the three-dimensional coordinates of all hub mark points and arm mark points are calculated through the intrinsic parameter matrices of the two cameras, the rotation matrix between the two cameras, and the translation matrix.

[0021] Furthermore, the method further comprises:

[0022] According to the relative position information of the known hub mark point, the three-dimensional information of the calculated three-dimensional coordinate point of the hub mark point is verified. If the error is less than 1‰, S4 is performed, otherwise the image containing the hub mark and the arm mark is re-collected.

[0023] Furthermore, S4 is specifically:

[0024] The three-dimensional coordinates of the hub marking point are fitted into a plane calculation normal vector. The three-dimensional coordinates of the same marking point on the four arms in different orientations are respectively fitted into a plane calculation normal vector to obtain five plane calculation normal vectors. The angle between all plane calculation normal vectors and the plane normal vector of the static hub marking point is calculated as the rotor disc inclination angle. The rotor disc inclination angle includes: the shaft inclination angle and the disc inclination angles of the four arms.

[0025] The invention has the advantages of simplicity, low cost and high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic flow chart of a method for measuring the inclination angle of a rotor blade disc based on binocular vision provided in an embodiment of the present invention;

[0027] Figure 2 To adjust the camera vertically downward to ensure that the rotor hub and the reflective target markings on the support arm are photographed within ±15°. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0029] Firstly, a binocular stereo vision camera and an LED light source are set up above the model rotor test bench, and the camera shoots the reflective targets pasted on the rotor hub and the support arm vertically downward; secondly, a photoelectric sensor is installed opposite to the rotating axis of the model rotor test bench, and the TOP signal generated by the photoelectric sensor is used to identify the azimuth and trigger the camera to shoot at equal azimuth intervals; thirdly, the camera is calibrated by identifying the reflective target pasted on the rotor hub to obtain the calibration coefficient; finally, the physical coordinates of the center of the reflective target are calculated according to the pixel coordinates of the center of the reflective target on the support arm and the calibration coefficient, and the normal vector is calculated by fitting the plane with the physical coordinates of the center.

[0030] Specifically, the process is as follows Figure 1 , a method for measuring the inclination angle of a rotor blade disc based on binocular vision, the method comprising:

[0031] S1, a binocular stereo vision camera and an LED light source are set up above the model rotor test bench, and the camera shoots the reflective target attached to the rotor hub and the support arm vertically downward; a photoelectric sensor is installed facing the rotating axis of the model rotor test bench, and the TOP signal generated by the photoelectric sensor is used to trigger the camera and identify the orientation of the camera shooting image;

[0032] S2, identifying the reflective target pasted on the rotor hub in the image captured by the camera, and combining the relative position information of the known hub marking points to calibrate the binocular camera to obtain a calibration matrix;

[0033] S3, determining the three-dimensional coordinates of all hub marking points and arm marking points;

[0034] S4, the three-dimensional coordinates of the hub mark point are fitted into a plane to calculate the normal vector, and the three-dimensional coordinates of the same mark point on the four arms at different orientations are respectively fitted into a plane to calculate the normal vector to obtain the rotor disc inclination angle.

[0035] S1 is specifically as follows: two high-speed cameras and an LED light source are set up vertically downward 2m above the model rotor test bench; a photoelectric sensor is installed facing the rotation axis of the test bench, and the photoelectric sensor is connected to the high-speed camera and the LED light source through a frequency doubling synchronizer.

[0036] The focal length of the camera lens is 50mm, and the field of view size is 1m*0.625m; the camera image resolution is set to 2048*1280pixel and the exposure time is 30 microseconds, the frequency synchronizer is set to 120 times, and the camera external trigger mode is turned on to collect images.

[0037] Before S2, the method also includes: adjusting the camera vertically downward to ensure that the rotor hub and the reflective target attached to the support arm are photographed within ±15°.

[0038] S2 specifically includes: when the rotation speed is 50 rpm in the hovering state, multiple images containing hub marking points are collected at each rotor shaft inclination angle and phase;

[0039] The YOLOv5 network model is used to identify and distinguish the arm markers and hub markers in the image, and the EDCircles circle detection method is used to detect the center coordinates of all markers.

[0040] Using the relative position information of the hub marking points obtained in the image and the known hub marking points, a physical template is established in the form of polar coordinates to calibrate the binocular camera. The intrinsic parameter matrices of the two cameras, the rotation matrix and the translation matrix between the two cameras are obtained as the calibration matrix.

[0041] S3 is specifically as follows: under a certain test state, an image containing a hub mark and an arm mark is collected, and the three-dimensional coordinates of all hub mark points and arm mark points are calculated through the intrinsic parameter matrices of the two cameras, the rotation matrix between the two cameras, and the translation matrix.

[0042] The method further includes: verifying the three-dimensional information of the three-dimensional coordinate point of the calculated hub mark point according to the relative position information of the known hub mark point, and if the error is less than 1‰, performing S4, otherwise re-collecting the image including the hub mark and the arm mark.

[0043] S4 is specifically as follows: the three-dimensional coordinates of the hub marking point are fitted into a plane calculation normal vector, and the three-dimensional coordinates of the same marking point on the four arms in different orientations are respectively fitted into a plane calculation normal vector to obtain five plane calculation normal vectors, and the angles between all plane calculation normal vectors and the plane normal vector of the static hub marking point are calculated as the rotor disc inclination angle. The rotor disc inclination angle includes: the shaft inclination angle and the disc inclination angles of the four arms.

[0044] Example

[0045] (1) Two high-speed cameras (FASTCAM MiniWX100type 80K-M-8GB) and an LED light source were set up vertically downward 2 m above the model rotor test bench. A photoelectric sensor was installed opposite the test bench's rotation axis and connected to the high-speed camera and LED light source through a frequency multiplier synchronizer.

[0046] (2) The focal length of the camera lens is 50 mm, and the field of view is 1 m*0.625 m. The camera image resolution is set to 2048*1280 pixels and the exposure time is set to 30 microseconds. The frequency synchronizer is set to 120 times, and the camera external trigger mode is turned on to collect images.

[0047] (3) Adjust the camera to point vertically downward to ensure that the rotor hub and the reflective target on the arm are captured within ±15°. The marking format is as follows: Figure 2 shown.

[0048] (4) The YOLOv5 network model is used to identify and distinguish the branch arm markers and the hub markers, and the EDCircles circle detection method is used to detect the center coordinates of all circular markers. i ,yb i ), hub (xc j ,yc j ).

[0049] (5) In the hovering state with a rotation speed of 50 rpm, 10 images containing hub marker points are collected at each rotor shaft inclination angle (±15°, ±10°, ±5°, ±2°) and other phases. The collected images are used to establish a physical template in polar coordinate form to calibrate the camera, and the intrinsic parameter matrices of the two cameras, the rotation matrix and the translation matrix between the two cameras are obtained respectively.

[0050] (6) Under a certain test state, an image containing the hub mark and the arm mark is collected, and the three-dimensional coordinates of all the hub mark points and the arm mark points are calculated through the intrinsic parameter matrices of the two cameras, the rotation matrix between the two cameras, and the translation matrix.

[0051] (7) Based on the relative position information of the known hub mark point, verify the three-dimensional information of the calculated three-dimensional coordinate point of the hub mark point. If the error is less than 1‰, proceed to step 8; otherwise, change the test state and proceed to step 6.

[0052] (8) The three-dimensional coordinates of the hub marking point are fitted into a plane calculation normal vector. The three-dimensional coordinates of the same marking point on the four arms at different orientations are respectively fitted into a plane calculation normal vector to obtain five plane calculation normal vectors. The angles between all the plane calculation normal vectors and the plane normal vector of the static hub marking point are calculated as the rotor disc inclination angle. The rotor disc inclination angle includes: the shaft inclination angle and the disc inclination angles of the four arms.

Claims

1. A method for measuring the inclination angle of a rotor blade disc based on binocular vision, characterized in that: The method comprises: S1, a binocular stereo vision camera and an LED light source are set up above the model rotor test bench, and the camera shoots the reflective target attached to the rotor hub and the support arm vertically downward; a photoelectric sensor is installed facing the rotating axis of the model rotor test bench, and the TOP signal generated by the photoelectric sensor is used to trigger the camera and identify the orientation of the camera shooting image; S2, identifying the reflective target pasted on the rotor hub in the image captured by the camera, and combining the relative position information of the known hub marking points to calibrate the binocular camera to obtain a calibration matrix; S3, determining the three-dimensional coordinates of all hub marking points and arm marking points; S4, the three-dimensional coordinates of the hub mark point are fitted into a plane to calculate the normal vector, and the three-dimensional coordinates of the same mark point on the four arms at different orientations are respectively fitted into a plane to calculate the normal vector to obtain the rotor disc inclination angle.

2. The method for measuring the rotor blade inclination angle based on binocular vision according to claim 1, characterized in that: S1 is specifically: Two high-speed cameras and an LED light source are set up vertically downward 2m above the model rotor test bench; a photoelectric sensor is installed opposite to the rotating axis of the test bench, and the photoelectric sensor is connected to the high-speed camera and the LED light source through a frequency doubling synchronizer.

3. The method for measuring the rotor blade inclination angle based on binocular vision according to claim 2, characterized in that: The focal length of the camera lens is 50mm, and the field of view size is 1m*0.625m; the camera image resolution is set to 2048*1280pixel and the exposure time is 30 microseconds, the frequency synchronizer is set to 120 times, and the camera external trigger mode is turned on to collect images.

4. The method for measuring rotor blade inclination angle based on binocular vision according to claim 2, characterized in that: Before S2, the method further includes: Adjust the camera vertically downward to ensure that the rotor hub and the reflective target attached to the arm are captured within ±15°.

5. The method for measuring rotor blade inclination angle based on binocular vision according to claim 4, characterized in that: S2 is specifically: In the hovering state with a speed of 50 rpm, multiple images containing hub markers are collected at each rotor shaft inclination phase; The YOLOv5 network model is used to identify and distinguish the arm markers and hub markers in the image, and the EDCircles circle detection method is used to detect the center coordinates of all markers. Using the relative position information of the hub marking points obtained in the image and the known hub marking points, a physical template is established in the form of polar coordinates to calibrate the binocular camera. The intrinsic parameter matrices of the two cameras, the rotation matrix and the translation matrix between the two cameras are obtained as the calibration matrix.

6. The method for measuring rotor blade inclination angle based on binocular vision according to claim 4, characterized in that: S3 is specifically: Under a certain test state, images containing hub marks and arm marks are collected, and the three-dimensional coordinates of all hub mark points and arm mark points are calculated through the intrinsic parameter matrices of the two cameras, the rotation matrix between the two cameras, and the translation matrix.

7. The method for measuring rotor blade inclination angle based on binocular vision according to claim 6, characterized in that: The method further comprises: According to the relative position information of the known hub mark point, the three-dimensional information of the calculated three-dimensional coordinate point of the hub mark point is verified. If the error is less than 1‰, S4 is performed, otherwise the image containing the hub mark and the arm mark is re-collected.

8. The method for measuring rotor blade inclination angle based on binocular vision according to claim 6, characterized in that: S4 is specifically: The three-dimensional coordinates of the hub marking point are fitted into a plane calculation normal vector. The three-dimensional coordinates of the same marking point on the four arms in different orientations are respectively fitted into a plane calculation normal vector to obtain five plane calculation normal vectors. The angle between all plane calculation normal vectors and the plane normal vector of the static hub marking point is calculated as the rotor disc inclination angle. The rotor disc inclination angle includes: the shaft inclination angle and the disc inclination angles of the four arms.

Images