Quaternion surface light field wheel alignment parameter detection system calibration method and system

Abstract

The application discloses a four-element number surface light field wheel positioning parameter detection system calibration method and system, and aims to solve the problem of four-element number surface light field wheel positioning parameter detection system calibration. The four-element number surface light field wheel positioning parameter detection system calibration method mainly comprises steps of image acquisition, surface light field solving and conversion matrix solving. The four-element number surface light field wheel positioning parameter detection system calibration system mainly comprises a top RGB-D camera (1), a right RGB-D camera (2), a right camera support (3), a surface laser (4), a surface laser support (5), a non-texture target plate (6), a target plate support (7), a left camera support (8), a left RGB-D camera (9) and a gantry (10). The four-element number surface light field wheel positioning parameter detection system calibration method and system can be used for large-scale space detection, has no calculation redundancy and stable performance.

Description

Technical Field

[0001] This invention relates to a calibration method and calibration equipment for measuring devices in the field of automotive testing; more specifically, it is a calibration method and system for a wheel alignment parameter detection system based on a quaternion surface light field. Background Technology

[0002] Vehicle inspection is a crucial research area for assessing vehicle passability and driving safety. Visual inspection, with its advantages of being non-contact, low-cost, and highly accurate, has become a key technology for improving the efficiency and accuracy of vehicle inspection in recent years. Machine vision can be used to detect vehicle wheel alignment parameters, wheelbase differences, vehicle shape, multi-axis axle offset angles, vehicle type recognition, and overload detection. Existing multi-RGB-D camera calibration methods mainly involve using point clouds with significant overlap acquired by the cameras to match corresponding feature points to obtain the pose relationship between two cameras; or using calibration objects such as spheres, cubes, planes, or human joint length constraints to calibrate the positional relationship between two or more cameras. These calibration systems require a large common field of view between the cameras and a small distance between them. However, the accuracy of point clouds acquired by RGB-D cameras at long distances is limited, making it difficult to extract corresponding feature points for accurate matching. This results in the inaccuracy of the pose relationship between the two RGB-D cameras obtained through point matching being difficult to guarantee. In comparison, surface laser devices are lightweight and flexible in operation, and the laser plane projected over long distances is less prone to divergence. Therefore, calibration algorithms for two RGB-D cameras based on laser plane normal vector matching typically achieve high accuracy. This paper proposes a quaternion-based surface light field calibration method and system for wheel alignment parameter detection. Multiple long-distance cameras in a vehicle wheel alignment parameter detection system based on a global light field network are calibrated using independent cameras, surface lasers, and textureless target plates. By using a three-dimensional laser plane as a bridge between two cameras and solving for the pose between them using quaternions, the coordinate system of each camera in the wheel alignment parameter detection system can be unified. Summary of the Invention

[0003] This invention addresses the problem of global calibration among multiple RGB-D cameras during vehicle wheel alignment parameter detection. It proposes a calibration method for a vehicle wheel alignment parameter detection system that is lightweight, flexible in operation, simple in structure, and possesses high precision in global surface light field. The detection system consists of a large-scale detection field composed of three triangularly distributed RGB-D cameras. A three-dimensional laser plane and a textureless target plate serve as bridges between two cameras to calibrate the pose relationships between the three RGB-D cameras. Quaternions are used to solve for the poses between two cameras, thus unifying the coordinate systems of each camera in the wheel alignment parameter detection system.

[0004] Referring to the accompanying drawings, the present invention is implemented using the following technical solution:

[0005] The specific steps of the calibration method for the wheel alignment parameter detection system based on quaternion surface light field are as follows:

[0006] Step 1: Image acquisition for the calibration method of the wheel alignment parameter detection system based on quaternion surface light field:

[0007] Place the left and right camera brackets and the gantry on the ground. Fix the left and right RGB-D cameras to the left and right camera brackets respectively. Fix the top RGB-D camera to the gantry. The left and right RGB-D cameras have no common field of view. Place the surface lasers on the ground and turn them on. Move the textureless target plate into the field of view of the left and top RGB-D cameras. The left and top RGB-D cameras will capture multiple images of the projected laser lines where the laser planes emitted by the surface lasers intersect with the surface of the textureless target plate. Then move the textureless target plate into the field of view of the right and top RGB-D cameras. The right and top RGB-D cameras will capture multiple images of the projected laser lines where the laser planes emitted by the surface lasers intersect with the surface of the textureless target plate. Turn off all surface lasers.

[0008] Step 2: Solving the surface light field of the wheel alignment parameter detection system calibration method using quaternion surface light field:

[0009] Taking the calibration of the positional relationship between the right-side RGB-D camera and the top RGB-D camera as an example, firstly, the three-dimensional point cloud of the textureless target plate is fitted to obtain the three-dimensional planar coordinates in the infrared camera coordinate system of the top RGB-D camera. k represents the position where the textureless target plate is placed at the k-th position within the field of view of the top RGB-D camera. Based on the color image of the projected laser line intersecting the laser plane and the surface of the textureless target plate acquired by the top RGB-D camera, the color image coordinates of the center feature point of the laser strip formed by the intersection of the planar target and the laser plane emitted by the surface laser are extracted. p represents the p-th (p = 1, 2…P) laser plane, q represents the q-th (q = 1, 2…Q) feature point on the p-th laser stripe on the textureless target plate, and the coordinates of the center feature point of the laser stripe in the color camera coordinate system of the top RGB-D camera are: Based on the projection relationship between the color camera coordinate system of the top RGB-D camera and its projected image coordinate system, we can obtain...

[0010]

[0011] in The intrinsic parameters of the color camera in the top RGB-D camera are obtained from the above formula. Based on the central feature point of the laser strip On a planar target, it can be obtained

[0012]

[0013] in This is the transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera;

[0014] Based on the intrinsic parameters of the color camera in the top RGB-D camera. The transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera. The 3D point cloud plane of the textureless target plate in the 3D planar coordinates of the infrared camera coordinate system of the top RGB-D camera. The center feature point of the laser strip can be obtained. The coordinates in the infrared coordinate system of the top RGB-D camera are:

[0015]

[0016] The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera are fitted to a plane using the RANSAC algorithm to obtain the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera.

[0017] The planar laser remains in the same position, and the planar target is moved into the field of view of the right RGB-D camera. The intrinsic parameters of the color camera within the right RGB-D camera are then used to determine the target's position. The transformation matrix between the coordinate systems of the infrared camera and the color camera in the RGB-D camera on the right. The 3D point cloud plane of the textureless target plate in the 3D planar coordinates of the infrared camera coordinate system of the top RGB-D camera. The center feature point of the laser strip can be obtained. Coordinates in the infrared coordinate system of the RGB-D camera on the right

[0018]

[0019] The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera are fitted into a plane using the RANSAC algorithm, resulting in the coordinates of the laser plane in the infrared camera coordinate system of the right RGB-D camera.

[0020] Finally, the surface laser is placed between the top RGB-D camera and the left RGB-D camera. Similarly, the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera can be obtained. And the coordinates of the laser plane in the infrared coordinate system of the left-hand RGB-D camera.

[0021] Step 3: Solving the transformation matrix of the quaternion surface light field wheel alignment parameter detection system calibration method:

[0022] Taking the calibration of the positional relationship between the top RGB-D camera and the left RGB-D camera as an example, based on the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera obtained in the second step... And the coordinates of the laser plane in the infrared coordinate system of the left-hand RGB-D camera. The imaginary quaternions of the laser plane normal vectors in the top RGB-D camera coordinate system and the left RGB-D camera coordinate system of the detection system are respectively expressed as: The unit quaternion representation of the rotation relationship between the left RGB-D camera and the top RGB-D camera coordinate system is as follows: The rotational relationship between the two planes can be expressed as follows:

[0023]

[0024] Based on the plane normal vector obtained from the left RGB-D camera After rotation, it should be aligned with the plane normal vector obtained from the top RGB-D camera. With the same direction, establish an objective function to solve for q. t,l for

[0025]

[0026] Where P is the total number of laser planes;

[0027] The rotation quaternion between the left RGB-D camera and the top RGB-D camera coordinate systems of the vehicle body reference wheel alignment parameter detection system, calculated using the above formula, is:

[0028] Based on the coordinates of the laser plane in the infrared coordinate system of the left RGB-D camera obtained in the second step. And the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera. Taken in the detection system A feature point within, whose three-dimensional coordinates are Taken in the detection system A feature point within, whose three-dimensional coordinates are Based on the properties of coordinate system transformation between points and surfaces, we have:

[0029]

[0030]

[0031] in To detect the normal vector parameters of the laser plane in the RGB-D camera coordinate system on the left side of the system. To detect the normal vector parameters of the laser plane in the RGB-D camera coordinate system at the top of the system. Translation vector T between the left RGB-D camera and the top RGB-D camera coordinate system t,l R t,2l The rotation matrix can be expressed as a quaternion by solving for it.

[0032]

[0033] Based on the feature points obtained above Coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera The above can be obtained

[0034] G t T t,l =d t

[0035] in

[0036] Using the above formula and three or more sets of non-parallel laser planes, the translation vector T between the left RGB-D camera and the top RGB-D camera coordinate system can be calculated. t,l When there are more than three laser planes, the translation vector between the left RGB-D camera and the top RGB-D camera coordinate system is calculated using SVD decomposition according to the following formula.

[0037] T t,l =((G t ) T G t ) -1 (G t ) T d t

[0038] Similarly, by placing the surface laser between the top RGB-D camera and the right RGB-D camera, the rotation quaternion q between the top RGB-D camera and the right RGB-D camera can be obtained. t,r and translation vector T t,r .

[0039] The calibration system of the quaternion surface light field wheel alignment parameter detection system includes a top RGB-D camera 1, a right RGB-D camera 2, a right camera bracket 3, a surface laser 4, a surface laser bracket 5, a textureless target plate 6, a target plate bracket 7, a left camera bracket 8, a left RGB-D camera 9, and a gantry 10.

[0040] The left camera bracket 8, the right camera bracket 3, the surface laser bracket 5, the target plate bracket 7, and the gantry 10 are placed on the ground. The left RGB-D camera 9 and the right RGB-D camera 2 are fixedly connected to the left camera bracket 8 and the right camera bracket 3 respectively through the threaded holes at the bottom. The left RGB-D camera 9 and the right RGB-D camera 2 have no common field of view. The surface laser 4 is fixedly connected to the top of the surface laser bracket 5 through the threaded bolts. The connecting shaft at the bottom of the textureless target plate 6 is fixedly connected to the threaded hole at the top of the target plate bracket 7. The top RGB-D camera 1 is fixedly connected to the crossbeam of the gantry 10 through the threaded hole at the bottom through the threaded bolts.

[0041] The beneficial effects of this invention are:

[0042] (1) The method of the present invention addresses the global calibration problem of cameras without a common field of view. It uses an independent camera, a surface laser, and a textureless target plate to perform global calibration of multiple RGB-D cameras, thereby achieving effective fusion of point cloud data from multiple RGB-D cameras and meeting the position measurement needs of distant objects such as positioning parameters between multiple wheels.

[0043] (2) The method of the two RGB-D camera coordinate system proposed in this invention can be extended to the study of multiple RGB-D coordinate systems. Therefore, it has important research value for the study of vehicle wheel positioning, wheel positioning parameters, wheelbase difference, shape reconstruction and other related fields.

[0044] (3) The method of the present invention introduces a surface laser as a bridge between RGB-D cameras. Since the laser plane can achieve long-distance propagation, the projected laser plane is not easy to diverge. Therefore, the two RGB-D camera calibration algorithm based on laser plane normal vector matching has high accuracy.

[0045] (4) The method proposed in this invention uses quaternions to solve the pose between two cameras. This way of describing rotation avoids the singularity problem in the Euler angle representation of rotation vectors and does not lead to redundancy of degrees of freedom due to the use of three-dimensional orthogonal matrices to represent rotation. Therefore, using quaternions to describe the rotation transformation between the coordinate systems of two RGB-D cameras is suitable for describing large-angle rotations of the coordinate systems of two RGB-D cameras, thus achieving the unification of the coordinate system of the detection system.

[0046] (5) The calibration system of the present invention has a wide calibration range, reliable performance, simple device structure, easy operation and low cost, and overcomes the disadvantage of the existing multi-RGB-D camera calibration method which requires large fixed contact calibration equipment between cameras. Attached Figure Description

[0047] Figure 1 This is a flowchart of the calibration method for a wheel alignment parameter detection system using a quaternion surface light field;

[0048] Figure 2 It is an axonometric drawing of a wheel alignment parameter detection and calibration system for a quaternion surface light field;

[0049] Figure 3 This is an isometric view of the top RGB-D camera 1 in the quaternion surface light field wheel positioning parameter detection system;

[0050] Figure 4 This is an isometric view of the right-side camera bracket 3 in the quaternion surface light field wheel positioning parameter detection system;

[0051] Figure 5 This is an isometric view of the surface laser 4 in the calibration system of the quaternion surface light field wheel alignment parameter detection system;

[0052] Figure 6 It is an axonometric view of the textureless target plate 6 in the calibration system of the quaternion surface light field wheel alignment parameter detection system;

[0053] Figure 7 This is an axonometric view of the gantry 10 in the quaternion surface light field wheel positioning parameter detection system;

[0054] In the image: 1. Top RGB-D camera, 2. Right RGB-D camera, 3. Right camera bracket, 4. Surface laser, 5. Surface laser bracket, 6. Textureless target plate, 7. Target plate bracket, 8. Left camera bracket, 9. Left RGB-D camera, 10. Gantry. Detailed Implementation

[0055] The invention will now be described in further detail with reference to the accompanying drawings:

[0056] See Figure 1 The calibration method for the wheel alignment parameter detection system of quaternion surface light field can be divided into the following three steps:

[0057] Step 1: Image acquisition for the calibration method of the wheel alignment parameter detection system based on quaternion surface light field:

[0058] Place the left camera bracket 8, the right camera bracket 3, and the gantry 10 on the ground. Fix the left RGB-D camera 9 and the right RGB-D camera 2 to the left camera bracket 8 and the right camera bracket 3 respectively. Fix the top RGB-D camera 1 to the gantry 10. The left RGB-D camera 9 and the right RGB-D camera 2 have no common field of view. Place the surface laser 4 on the ground and turn on the surface laser 4. Move the textureless target plate 6 into the field of view of the left RGB-D camera 9 and the top RGB-D camera 1. The left RGB-D camera 9 and the top RGB-D camera 1 will capture multiple images of the projected laser lines where the laser plane emitted by the surface laser 4 intersects with the surface of the textureless target plate 6. Then move the textureless target plate 6 into the field of view of the right RGB-D camera 2 and the top RGB-D camera 1. The right RGB-D camera 2 and the top RGB-D camera 1 will capture multiple images of the projected laser lines where the laser plane emitted by the surface laser 4 intersects with the surface of the textureless target plate 6. Turn off all surface lasers 4.

[0059] Step 2: Solving the surface light field of the wheel alignment parameter detection system calibration method using quaternion surface light field:

[0060] Taking the calibration of the positional relationship between the right-side RGB-D camera 2 and the top RGB-D camera 1 as an example, firstly, the three-dimensional point cloud of the textureless target plate 6 is fitted to obtain the three-dimensional planar coordinates in the infrared camera coordinate system of the top RGB-D camera 1. k represents the k-th position of the textureless target plate 6 within the field of view of the top RGB-D camera 1. Based on the color image of the projected laser line where the laser plane intersects the surface of the textureless target plate 6, captured by the top RGB-D camera 1, the color image coordinates of the center feature point of the laser strip formed by the intersection of the planar target and the laser plane emitted by the surface laser 4 are extracted. p represents the p-th (p = 1, 2…P) laser plane, q represents the q-th (q = 1, 2…Q) feature point on the p-th laser stripe on the textureless target plate 6, and the coordinates of the center feature point of the laser stripe in the color camera coordinate system of the top RGB-D camera 1 are: Based on the projection relationship between the color camera coordinate system of the top RGB-D camera 1 and its projected image coordinate system, we can obtain...

[0061]

[0062] in The intrinsic parameters of the color camera in the top RGB-D camera 1 can be obtained from the above formula. Based on the central feature point of the laser strip On a planar target, it can be obtained

[0063]

[0064] in This is the transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera 1;

[0065] Based on the intrinsic parameters of the color camera in the top RGB-D camera 1 The transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera 1 The 3D point cloud plane of the textureless target plate 6 in the 3D planar coordinates of the infrared camera coordinate system of the top RGB-D camera 1 The center feature point of the laser strip can be obtained. The coordinates in the infrared coordinate system of the top RGB-D camera 1 are:

[0066]

[0067] The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera 1 are fitted to a plane using the RANSAC algorithm, thus obtaining the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera 1.

[0068] The planar laser 4 remains in the same position, and the planar target is moved into the field of view of the right RGB-D camera 2, based on the intrinsic parameters of the color camera in the right RGB-D camera 2. The transformation matrix between the coordinate systems of the infrared camera and the color camera in the RGB-D camera 2 on the right. The 3D point cloud plane of the textureless target plate 6 in the 3D planar coordinates of the infrared camera coordinate system of the top RGB-D camera 1 The center feature point of the laser strip can be obtained. Coordinates in the infrared coordinate system of the RGB-D camera on the right

[0069]

[0070] The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera 2 are fitted into a plane using the RANSAC algorithm, resulting in the coordinates of the laser plane in the infrared camera coordinate system of the right RGB-D camera 2.

[0071] Finally, the surface laser 4 is placed between the top RGB-D camera 1 and the left RGB-D camera 9. Similarly, the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera 1 can be obtained. And the coordinates of the laser plane in the infrared coordinate system of the RGB-D camera on the left.

[0072] Step 3: Solving the transformation matrix of the quaternion surface light field wheel alignment parameter detection system calibration method:

[0073] Taking the calibration of the positional relationship between the top RGB-D camera 1 and the left RGB-D camera 9 as an example, the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera 1 obtained in the second step are used as an example. And the coordinates of the laser plane in the infrared coordinate system of the RGB-D camera on the left. The imaginary quaternions of the laser plane normal vectors in the top RGB-D camera 1 coordinate system and the left RGB-D camera 9 coordinate system of the detection system are respectively expressed as: The unit quaternion representation of the rotation relationship between the coordinate systems of the left RGB-D camera 9 and the top RGB-D camera 1 is as follows: The rotational relationship between the two planes can be expressed as:

[0074]

[0075] Based on the plane normal vector obtained from the left RGB-D camera 9 After rotation, it should be aligned with the plane normal vector obtained by the top RGB-D camera 1. With the same direction, establish an objective function to solve for q. t,l for

[0076]

[0077] Where P is the total number of laser planes;

[0078] The rotation quaternion between the coordinate systems of the left RGB-D camera 9 and the top RGB-D camera 1 in the vehicle body reference wheel alignment parameter detection system is calculated using the above formula.

[0079] Based on the coordinates of the laser plane in the infrared coordinate system of the left RGB-D camera obtained in the second step. And the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera 1 Taken in the detection system A feature point within, whose three-dimensional coordinates are Taken in the detection system A feature point within, whose three-dimensional coordinates are Based on the properties of coordinate system transformation between points and surfaces, we have:

[0080]

[0081]

[0082] in To detect the normal vector parameters of the laser plane in the 9-coordinate system of the RGB-D camera on the left side of the system. To detect the normal vector parameters of the laser plane in the coordinate system of the RGB-D camera at the top of the system. Translation vector T between the coordinate systems of the left RGB-D camera 9 and the top RGB-D camera 1 t,l R t,2l The rotation matrix can be expressed as a quaternion by solving for it.

[0083]

[0084] Based on the feature points obtained above Coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera 1 The above can be obtained

[0085] G t T t,l =d t

[0086] in

[0087] Using the above formula and three or more sets of non-parallel laser planes, the translation vector T between the coordinate system of the left RGB-D camera 9 and the top RGB-D camera 1 can be calculated. t,l When there are more than three laser planes, the translation vector between the coordinate systems of the left RGB-D camera 9 and the top RGB-D camera 1 is calculated using SVD decomposition according to the following formula.

[0088] T t,l =((G t ) T G t ) -1 (G t ) T d t

[0089] Similarly, by placing the surface laser 4 between the top RGB-D camera 1 and the right RGB-D camera 2, the rotation quaternion q between the top RGB-D camera 1 and the right RGB-D camera 2 can be obtained. t,r and translation vector T t,r .

[0090] See Figures 2 to 7 The calibration system of the quaternion surface light field wheel positioning parameter detection system includes a top RGB-D camera 1, a right RGB-D camera 2, a right camera bracket 3, a surface laser 4, a surface laser bracket 5, a textureless target plate 6, a target plate bracket 7, a left camera bracket 8, a left RGB-D camera 9, and a gantry 10.

[0091] The left camera bracket 8, the right camera bracket 3, the surface laser bracket 5, the target plate bracket 7, and the gantry 10 are placed on the ground. The left RGB-D camera 9 and the right RGB-D camera 2 are fixedly connected to the left camera bracket 8 and the right camera bracket 3 respectively through the threaded holes at the bottom. The left RGB-D camera 9 and the right RGB-D camera 2 have no common field of view. The surface laser 4 is fixedly connected to the top of the surface laser bracket 5 through the threaded bolts. The connecting shaft at the bottom of the textureless target plate 6 is fixedly connected to the threaded hole at the top of the target plate bracket 7. The top RGB-D camera 1 is fixedly connected to the crossbeam of the gantry 10 through the threaded hole at the bottom through the threaded bolts.

Claims

1. A calibration method for a wheel alignment parameter detection system using a quaternion surface light field, characterized in that, The specific steps of the calibration method for the wheel positioning parameter detection system based on the quaternion surface light field are as follows: Step 1: Image acquisition for the calibration method of the wheel alignment parameter detection system based on quaternion surface light field: Place the left camera bracket (8), right camera bracket (3), and gantry (10) on the ground. Fix the left RGB-D camera (9) and right RGB-D camera (2) to the left camera bracket (8) and right camera bracket (3) respectively. Fix the top RGB-D camera (1) to the gantry (10). The left RGB-D camera (9) and right RGB-D camera (2) have no common field of view. Place the surface laser (4) on the ground and turn on the surface laser (4). Move the textureless target plate (6) into the left RGB-D camera (9) and top RGB-D camera (10). Within the field of view of camera (1), the left RGB-D camera (9) and the top RGB-D camera (1) acquire images of the projected laser lines where the laser planes emitted by the multi-face lasers (4) intersect with the surface of the textureless target plate (6); then the textureless target plate (6) is moved into the field of view of the right RGB-D camera (2) and the top RGB-D camera (1), and the right RGB-D camera (2) and the top RGB-D camera (1) acquire images of the projected laser lines where the laser planes emitted by the multi-face lasers (4) intersect with the surface of the textureless target plate (6), and all face lasers (4) are turned off; Step 2: Solving the surface light field of the wheel alignment parameter detection system calibration method using quaternion surface light field: Taking the calibration of the positional relationship between the right RGB-D camera (2) and the top RGB-D camera (1) as an example, firstly, the three-dimensional point cloud of the textureless target plate (6) is fitted to obtain the three-dimensional planar coordinates in the infrared camera coordinate system of the top RGB-D camera (1). k represents the k-th position of the textureless target plate (6) within the field of view of the top RGB-D camera (1). Based on the color image of the projected laser line where the laser plane intersects with the surface of the textureless target plate (6) acquired by the top RGB-D camera (1), the color image coordinates of the center feature point of the laser strip formed by the intersection of the planar target and the laser plane emitted by the surface laser (4) are extracted. p represents the p-th (p = 1, 2…P) laser plane, q represents the q-th (q = 1, 2…Q) feature point on the p-th laser stripe on the textureless target plate (6), and the coordinates of the center feature point of the laser stripe in the color camera coordinate system of the top RGB-D camera (1) are: Based on the projection relationship between the color camera coordinate system of the top RGB-D camera (1) and its projected image coordinate system, we can obtain... in The intrinsic parameters of the color camera in the top RGB-D camera (1) can be obtained from the above formula. Based on the central feature point of the laser strip On a planar target, it can be obtained in The transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera (1); Based on the intrinsic parameters of the color camera in the top RGB-D camera (1) The transformation matrix between the coordinate systems of the infrared camera and the color camera in the top RGB-D camera (1) The three-dimensional point cloud plane of the textureless target plate (6) in the three-dimensional plane coordinates of the infrared camera coordinate system of the top RGB-D camera (1) The center feature point of the laser strip can be obtained. The coordinates in the infrared coordinate system of the top RGB-D camera (1) are: The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera (1) are fitted into a plane using the RANSAC algorithm to obtain the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera (1). The surface laser (4) remains in the same position, and the planar target is moved into the field of view of the right RGB-D camera (2). Based on the intrinsic parameters of the color camera in the right RGB-D camera (2), The transformation matrix between the coordinate systems of the infrared camera and the color camera in the RGB-D camera (2) on the right. The three-dimensional point cloud plane of the textureless target plate (6) in the three-dimensional plane coordinates of the infrared camera coordinate system of the top RGB-D camera (1) The center feature point of the laser strip can be obtained. Coordinates in the infrared coordinate system of the RGB-D camera (2) on the right The laser strip center feature point obtained above The coordinates in the infrared coordinate system of the top RGB-D camera 2 are fitted into a plane using the RANSAC algorithm, and the coordinates of the laser plane in the infrared camera coordinate system of the right RGB-D camera (2) are obtained. Finally, the surface laser (4) is placed between the top RGB-D camera (1) and the left RGB-D camera (9). Similarly, the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera (1) can be obtained. And the coordinates of the laser plane in the infrared coordinate system of the left RGB-D camera (9) Step 3: Solving the transformation matrix of the quaternion surface light field wheel alignment parameter detection system calibration method: Taking the positional relationship calibration between the top RGB-D camera (1) and the left RGB-D camera (9) as an example, the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera (1) obtained in the second step are used as an example. And the coordinates of the laser plane in the infrared coordinate system of the left RGB-D camera (9) The imaginary quaternions of the laser plane normal vectors in the coordinate systems of the top RGB-D camera (1) and the left RGB-D camera (9) in the detection system are respectively expressed as: The unit quaternion representation of the rotation relationship between the coordinate systems of the left RGB-D camera (9) and the top RGB-D camera (1) is as follows: The rotational relationship between the two planes can be expressed as: Based on the plane normal vector obtained from the left RGB-D camera (9) After rotation, it should be aligned with the plane normal vector obtained by the top RGB-D camera (1). With the same direction, establish an objective function to solve for q. t,l for Where P is the total number of laser planes; The rotation quaternion between the coordinate systems of the left RGB-D camera (9) and the top RGB-D camera (1) of the vehicle body reference wheel alignment parameter detection system is calculated according to the above formula. Based on the coordinates of the laser plane in the infrared coordinate system of the left RGB-D camera (9) obtained in the second step And the coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera (1) Taken in the detection system A feature point within, whose three-dimensional coordinates are Taken in the detection system A feature point within, whose three-dimensional coordinates are Based on the properties of coordinate system transformation between points and surfaces, we have: in To detect the normal vector parameters of the laser plane in the coordinate system of the RGB-D camera (9) on the left side of the system. To detect the normal vector parameters of the laser plane in the coordinate system of the RGB-D camera (1) at the top of the system. Translation vector T between the coordinate systems of the left RGB-D camera (9) and the top RGB-D camera (1) t,l R t,2l The rotation matrix can be expressed as a quaternion by solving for it. Based on the feature points obtained above Coordinates of the laser plane in the infrared coordinate system of the top RGB-D camera (1) The above can be obtained G t T t,l =d t in Using the above formula and three or more sets of non-parallel laser planes, the translation vector T between the coordinate system of the left RGB-D camera (9) and the top RGB-D camera (1) can be calculated. t,l When there are more than three laser planes, the translation vector between the coordinate systems of the left RGB-D camera (9) and the top RGB-D camera (1) is calculated using SVD decomposition according to the following formula. T t,l =((G t ) T G t ) -1 (G t ) T d t Similarly, by placing the surface laser (4) between the top RGB-D camera (1) and the right RGB-D camera (2), the rotation quaternion q between the top RGB-D camera (1) and the right RGB-D camera (2) can be obtained. t,r and translation vector T t,r .

2. The calibration system for the calibration method of the wheel positioning parameter detection system of the quaternion surface light field according to claim 1, characterized in that, The calibration system of the quaternion surface light field wheel positioning parameter detection system calibration method includes a top RGB-D camera (1), a right RGB-D camera (2), a right camera bracket (3), a surface laser (4), a surface laser bracket (5), a textureless target plate (6), a target plate bracket (7), a left camera bracket (8), a left RGB-D camera (9), and a gantry (10); The left camera bracket (8), the right camera bracket (3), the surface laser bracket (5), the target plate bracket (7), and the gantry (10) are placed on the ground. The left RGB-D camera (9) and the right RGB-D camera (2) are fixedly connected to the top bolts of the left camera bracket and the right camera bracket (3) respectively through the threaded holes at the bottom. The left RGB-D camera (9) and the right RGB-D camera (2) have no common field of view. The surface laser (4) is fixedly connected to the top bolt of the surface laser bracket (5). The connecting shaft at the bottom of the textureless target plate (6) is fixedly connected to the threaded hole at the top of the target plate bracket (7). The top RGB-D camera (1) is fixedly connected to the bolt of the crossbeam of the gantry (10) through the threaded hole at the bottom.

Images