Camera-lantern space pose alignment method based on four LED array and turntable

Abstract

This invention relates to a camera-collimator spatial pose alignment method, specifically a method based on a four-LED array and a turntable, which solves the problems of lack of quantization and high-precision full-attitude alignment in existing technologies. The method specifically includes: Step 1, setting a square four-LED array at the light outlet of the collimator, and placing the camera on the turntable via a six-DOF adjustment stage; Step 2, controlling the rotation of the turntable and acquiring a first image of the four-LED array; Step 3, calculating and obtaining multiple sets of ordered coordinates based on the first image; Step 4, calculating the rotation matrix, Euler angles, and translation vector; Step 5, calculating the six-DOF deviation vector of the turntable axis; Step 6, calculating the translation adjustment and rotation adjustment; Step 7, adjusting the camera pose based on the translation and rotation adjustment to complete the camera-collimator spatial pose alignment. This invention makes the alignment reference quantifiable, improving alignment accuracy.

Description

Technical Field

[0001] This invention relates to a camera-collimator spatial pose alignment method, specifically a camera-collimator spatial pose alignment method based on a four-LED array and a turntable. Background Technology

[0002] The field of view of a camera is an important technical indicator characterizing the imaging range of a camera. Existing methods for measuring the field of view typically employ a turntable approach: the camera mounting surface is fixed to the turntable plane, the object-side node of the camera's optical system coincides with the rotation center of the turntable, and the collimator is aligned so that the star point aperture placed at the focal plane of the collimator images onto the camera target surface; then the turntable is rotated so that the star point image moves from one side of the image to the other, and the angle through which the turntable rotates is the field of view.

[0003] Chinese invention patent CN103528559A discloses a camera field-of-view calibration method and apparatus. Its hardware configuration includes: attaching a characteristic circular pattern to the center of a target plate; fixing the camera to an electric translation stage via a five-dimensional adjustment frame, with the electric translation stage's running direction perpendicular to the target plate plane. Specific operating steps include: adjusting the five-dimensional adjustment frame so that the circular pattern at the center of the target plate is imaged at the center of the image captured by the camera, while ensuring symmetrical imaging of the target plate vertically and horizontally; recording the camera position 'a' and the horizontal field-of-view width 'P' at this point. a Subsequently, the electric translation stage moved the camera to position b, and the five-dimensional adjustment frame was readjusted to re-image the circular shape at the center of the image while maintaining the vertical and horizontal symmetry of the target plate. Position b and the horizontal field of view width P were recorded. b The distance ΔL to be moved is calculated using a formula based on the principle of triangle similarity, and the field of view is expressed as... .

[0004] However, this solution has the following technical drawbacks:

[0005] Defect 1: Alignment relies on manual visual judgment and lacks quantitative benchmarks.

[0006] In this patent, the claims of "imaging a circular graphic at the center of the image" and "the target plate imaging being symmetrical vertically and horizontally" both rely on the operator's visual observation. The entire specification does not provide quantitative criteria for "imaging at the center of the image." The operator can only rely on experience to repeatedly adjust the five-dimensional adjustment frame until the conditions are subjectively determined by the naked eye, resulting in a highly subjective, low-precision, and poor-repeatable alignment process.

[0007] Defect 2: It only performs coarse two-dimensional position and coarse attitude adjustments, lacking high-precision full attitude alignment.

[0008] The "image center" alignment in this patent only ensures the position of feature points on the two-dimensional plane of the image, while the "vertical and horizontal symmetry" adjustment can only correct coarse attitude deviations in the pitch and yaw directions, making the target plane normal roughly parallel to the optical axis. However, this scheme does not involve attitude adjustment in the roll direction; its adjustment target is only "image symmetry" of the target, rather than making the camera detection surface and the target plane achieve precise spatial parallelism, and it lacks quantitative numerical feedback of angular deviations. More importantly, when there are small pitch or yaw angle deviations in the camera, the two conditions of "circular shape located at the image center" and "target plane image symmetry" can still be met simultaneously, but this deviation cannot be detected and quantified, and directly substituting it into the calculation formula will introduce systematic errors.

[0009] Defect 3: Limited application scenarios.

[0010] The patent title and background art explicitly limit it to a "near-range imaging camera field-of-view calibration" method. Its calibration relies on the physical dimensions of the target plate and the movement distance of the motorized translation stage, making it only applicable to near-range scenarios. For field-of-view measurement scenarios using collimators to simulate targets at infinity (such as star sensors, telephoto lenses, and space cameras), the hardware configuration and operation procedures of this method are not suitable. Furthermore, this method calibrates the line field of view based on the principle of triangle similarity, and its calculation formula implicitly includes small-angle approximation conditions, leading to a significant decrease in accuracy at longer distances or larger field-of-view angles.

[0011] Defect 4: The alignment process is an open-loop operation and lacks an error convergence mechanism.

[0012] In this patent, the operator stops adjusting once they visually determine that the alignment is correct. The entire adjustment process is an open-loop operation: no verification measurement is performed after adjustment, making it impossible to confirm whether the actual positional deviation truly falls within the allowable error range. When the adjustment frame has mechanical backlash or the operator makes a misjudgment, the adjustment result is neither traceable nor verifiable. Currently, there is no solution to construct the alignment process as a closed-loop iterative control system of "measurement → adjustment → re-measurement" to achieve automatic convergence of alignment errors.

[0013] Furthermore, camera alignment requires knowledge of the camera's pose relative to the world coordinate system, and camera intrinsic parameters (including normalized focal length) are also required. and the main point Corner coordinates are essential parameters for calculating relative pose and must be calibrated in advance. The current mainstream method in the industrial field is the "Zhang Zhengyou checkerboard calibration method," which uses a checkerboard calibration board to capture multiple images from different orientations, extracts corner coordinates, and solves for intrinsic parameters through homography matrix constraints. There are also self-calibration methods based on LED arrays, which use LED arrays with known spacing to capture multiple images from different orientations to complete the calibration. However, all of the above methods require a dedicated process separate from the alignment process and cannot be integrated into the alignment process. Summary of the Invention

[0014] This invention aims to solve the technical problems in the prior art, such as lack of quantization in alignment, lack of high-precision full-attitude alignment, limited application scenarios, lack of error convergence mechanism in the alignment process, and the need for pre-calibration of the camera. It proposes a camera-parallel light tube spatial pose alignment method based on a four-LED array and a turntable.

[0015] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0016] A camera-collimator spatial pose alignment method based on a four-LED array and a turntable, characterized by the following steps:

[0017] Step 1: Set up a square four-LED array at the light outlet of the collimator, with the array center on the optical axis of the collimator and the array surface perpendicular to the optical axis. Set the camera on a six-degree-of-freedom adjustment stage and place the six-degree-of-freedom adjustment stage on a turntable. Adjust the six-degree-of-freedom adjustment stage so that all four LED arrays are in the camera's field of view.

[0018] Step 2: Control the turntable to rotate multiple times. After each rotation, the camera captures one frame of the first image of the four-LED array, resulting in multiple frames of the first image.

[0019] Step 3: Calculate the sub-pixel coordinates and polar angles of the four LEDs in the first image of each frame, and assign ordered numbers to the four sub-pixel coordinates according to the size of the polar angles to obtain multiple sets of ordered coordinates;

[0020] Step 4: Establish and solve an overdetermined system of equations based on multiple sets of ordered coordinates to obtain the rotation matrix, Euler angles and translation vector from the world coordinate system of the collimator to the camera coordinate system when the turntable axis is at 0°.

[0021] Step 5: Calculate the six-degree-of-freedom deviation vector of the turntable axis using the Euler angles of the rotation matrix of the collimator relative to the camera and the translation vector when the turntable axis is at 0°.

[0022] Step 6: Calculate the translation and rotation adjustments using the six-degree-of-freedom deviation vector;

[0023] Step 7: Adjust the camera pose based on the translation and rotation adjustments to complete the camera-parallel light tube spatial pose alignment.

[0024] Further, step 1 specifically involves setting up a square four-LED array on the collimator; the side length of the square is 2a, a>0, the center of the square is located on the optical axis of the collimator, its horizontal direction is horizontal and its vertical direction is vertical; the camera is set on a six-degree-of-freedom adjustment stage, and the six-degree-of-freedom adjustment stage is set on a turntable, and a frame of array image of the four-LED array is acquired by the camera, and the six-degree-of-freedom adjustment stage is adjusted according to the array image so that the four LEDs are located in the four quadrant regions of the array image respectively, so that the four-LED array is completely in the camera's field of view.

[0025] Further, step 2 specifically involves controlling the turntable to rotate N times, where N ≥ 3, with each rotation angle being θ. k , k∈[1,N], θ max / 10≥θ k >0°, θ max The camera's horizontal maximum field of view is defined by the first image I of the four-LED array acquired after each rotation. k , obtain N frames of the first image I k .

[0026] Further, step 3 specifically involves processing each frame of the first image I... k Perform the following processing to obtain N sets of ordered coordinates:

[0027] Step A: Calculate the first image I using the gray-scale centroid method. k Subpixel coordinates of each LED spot ( ):

[0028] ; ;

[0029] Among them, I mn,i For the first image I k The pixel grayscale value of the i-th LED spot region, m i and n i For the first image I k The column and row numbers of the pixels in the i-th LED spot region, i = 1, 2, 3, 4; the summation range covers the i-th LED spot region;

[0030] Step B, using four sub-pixel coordinates ( ) Calculate the first image I k Geometric center coordinates ( ):

[0031] ;

[0032] ;

[0033] Step C, using the geometric center coordinates ( Establish a polar coordinate system with the origin as the origin, and calculate the polar angles of the four LED spots. :

[0034] ;

[0035] Step D: Set the four sub-pixel coordinates ( Sort the coordinates in ascending order according to their corresponding polar angles and assign them sequential numbers 1, 2, 3, and 4 to obtain N sets of ordered coordinates.

[0036] Further, step 4 specifically involves establishing an overdetermined system of equations based on the N sets of ordered coordinates, and solving it using the Gauss-Newton method or the Levenberg-Marquardt algorithm to obtain the Euler angles of the rotation matrix from the world coordinate system of the collimator to the camera coordinate system when the turntable axis is at 0°. , , Translation vector , , :

[0037] ;

[0038] in, The rotation matrix of the collimator relative to the camera when the turntable axis is at 0°; is the translation vector of the collimator relative to the camera when the turntable axis is at 0°; n is the unit vector of the axis direction. For rotation angle θ k The corresponding rotation matrix; I is a 3×3 identity matrix; T is the transpose; Let n be an antisymmetric matrix; and These are the rotation matrix and translation vector of the collimator relative to the camera when it is rotated to the kth time, respectively. Let i be the coordinates of the i-th LED in the world coordinate system. , , , ; q=(q x ,q y ,q z ) T f is the point on the axis of rotation closest to the origin of the world coordinate system. x f y These are the camera's normalized focal length along the x-axis and y-axis, respectively; , [] represents the principal point coordinates of the camera; the subscripts x, y, and z represent the components of the column vector on the corresponding axes of the camera coordinate system.

[0039] Further, step 5 specifically involves using the Euler angle rotation matrix. , , Translation vector , , Calculate the six-degree-of-freedom deviation vector of the turntable axis. :

[0040] ;

[0041] in, ΔX, ΔY, and ΔZ represent the deviations along the x, y, and z axes of the camera coordinate system, respectively; ΔZ is only used as a monitoring value for axial distance and does not participate in alignment; L design This is a preset distance between the optical axis of the collimator and the principal point of the camera under ideal conditions. , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively. , , .

[0042] Further, step 6 specifically involves using the six-degree-of-freedom deviation vector... Calculate the translation and rotation adjustments:

[0043] Translation adjustment amount: ;

[0044] Rotation adjustment amount: .

[0045] Furthermore, it also includes:

[0046] Step 8: Verify that the camera-collimator spatial pose is aligned;

[0047] Step 8.1: Reacquire N frames of the second image of the four-LED array using the method in Step 2, and recalculate the six-degree-of-freedom deviation vector using the methods in Steps 3 to 5.

[0048] Step 8.2: Compare the absolute values ​​of each component in the recalculated six-degree-of-freedom deviation vector with the corresponding preset tolerance threshold. Determine whether the alignment is complete based on the comparison results. If yes, the alignment is considered complete; otherwise, proceed to step 8.3.

[0049] Step 8.3: Recalculate the translation and rotation adjustments using the recalculated six-degree-of-freedom deviation vector;

[0050] Step 8.4: Adjust the camera pose based on the recalculated translation and rotation adjustments, and return to step 8.1 until alignment is determined to be complete.

[0051] Further, step 8.1 specifically involves re-acquiring N frames of the second image of the four-LED array using the method in step 2, and recalculating the six-degree-of-freedom deviation vector using the methods in steps 3 to 5. :

[0052] ;

[0053] in, , , These are the deviations on the recalculated x, y, and z axes, respectively. It is used only as a monitoring value for axial distance and is not involved in alignment; , , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively.

[0054] Furthermore, it also includes:

[0055] Step 8.2 specifically involves recalculating the six-degree-of-freedom deviation vector. The absolute values ​​of each component and their corresponding preset tolerance thresholds , , , , Compare and determine if the following conditions are met:

[0056] ;

[0057] ;

[0058] ;

[0059] ;

[0060] ;

[0061] in, This is the preset tolerance threshold for the deviation on the x-axis. This is the preset tolerance threshold for the deviation on the y-axis. This is the preset tolerance threshold for pitch angle deviation. This is the preset tolerance threshold for yaw angle deviation. This is the preset tolerance threshold for roll angle deviation;

[0062] If all of the above conditions are met, the alignment is considered complete.

[0063] If any of the above conditions are not met, proceed to step 8.3;

[0064] Step 8.3 specifically involves recalculating the six-degree-of-freedom deviation vector. Recalculate the translation and rotation adjustments.

[0065] The beneficial effects of this invention are:

[0066] 1. Alignment reference can be quantified:

[0067] This invention transforms the alignment problem from a qualitative description (enabling a circular image to be imaged at the center of the image) into a quantitative calculation problem based on the coordinate set of four LED sub-pixels and the physical center of the image, thus eliminating the subjectivity and uncertainty of human visual judgment.

[0068] 2. Complete alignment of six degrees of freedom:

[0069] Unlike existing technologies that only focus on two-dimensional position alignment and coarse attitude correction (only pitch and yaw directions), this invention covers complete pose alignment with six degrees of freedom, including position (x / y / z translation) and attitude (pitch / yaw / roll rotation).

[0070] 3. Numerical output of deviation:

[0071] By establishing a 16-parameter overdetermined equation system, the specific values ​​of the deviation of each degree of freedom are calculated, providing a quantitative basis for precise adjustment. This is different from existing technologies that can only qualitatively determine "whether it is centered" but cannot know "how much it is off".

[0072] 4. High degree of automation:

[0073] This invention inverts the deviation and compensates it simultaneously through the adjustment stage. After following the steps of "acquisition → calculation → adjustment → convergence judgment → iterative adjustment", it will automatically converge to a fully aligned state, which significantly improves alignment efficiency and repeatability. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the structure of the collimator, the four-LED array, and the camera in an embodiment of a camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to the present invention.

[0075] The attached figures are labeled as follows:

[0076] 1. Collimator; 2. Four-LED array; 3. Camera; 4. Six-DOF adjustment stage; 5. Turntable. Detailed Implementation

[0077] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] This invention provides a camera-collimator spatial pose alignment method based on a four-LED array and a turntable, comprising the following steps:

[0079] Step 1, as follows Figure 1 As shown, a square four-LED array 2 is set at the light outlet of the collimator 1. The camera 3 is set on a six-degree-of-freedom adjustment stage 4, which is then set on a turntable 5. The six-degree-of-freedom adjustment stage 4 is adjusted so that the entire four-LED array 2 enters the field of view of the camera 3. Specifically:

[0080] A square four-LED array 2 is set at the light outlet of the collimator 1. The side length of the square is 2a, where a > 0. The center of the array is on the optical axis of the collimator 1, and the array surface is perpendicular to the optical axis. Its horizontal direction is horizontal, and its vertical direction is vertical. A camera 3 is set on a six-degree-of-freedom adjustment stage 4, which is set on a turntable 5. A frame of the four-LED array 2 is captured by the camera 3. The six-degree-of-freedom adjustment stage 4 is adjusted according to the array image so that the four LEDs are located in the four quadrant regions of the array image (one each in the upper left, upper right, lower left, and lower right), so that all four LEDs are in the field of view of the camera 3. The four LEDs are of the same specification and are not individually numbered. The turntable 5 can rotate around its axis to a known angle (the angle can be read from the turntable). The six-degree-of-freedom adjustment stage 4 on the turntable 5 is an electric adjustment stage with electric adjustment capability of six degrees of freedom: X / Y / Z translation and pitch / yaw / roll rotation. Each axis is equipped with a precision grating ruler or encoder as a position feedback element. Both the six-degree-of-freedom adjustment stage 4 and the turntable 5 are controlled by a computer system, which includes an image acquisition and processing module, a parameter calculation module, and a servo control module. The servo control module sends adjustment commands to the drivers of each axis of the six-degree-of-freedom adjustment stage 4 based on the calculated deviation, and determines whether further iteration is needed based on feedback from newly acquired images. In this embodiment, the side length of the square is 2a = 100mm.

[0081] Step 2: Control the turntable 5 to rotate multiple times. After each rotation, the camera 3 captures a frame of the first image I of the four-LED array. k , to obtain multiple frames of the first image I k Specifically:

[0082] Control the turntable 5 to rotate N times, where N≥3, and the rotation angle is θ each time. k, k∈[1,N], θ max / 10≥θ k >0°, θ max To determine the maximum horizontal field of view of the camera, camera 3 acquires a first image I of the four-LED array after each rotation. k , obtain N frames of the first image I k In this embodiment, N=3, and the rotation angles of the turntable 5 are 0°, 5°, and -5°.

[0083] Step 3: Calculate the first image I of each frame. k The sub-pixel coordinates and polar angles of the four LEDs are determined, and the four sub-pixel coordinates are assigned ordered numbers according to the size of the polar angles, resulting in multiple sets of ordered coordinates; specifically:

[0084] For each frame, the first image I k Perform the following processing to obtain N sets of ordered coordinates:

[0085] Step A: Calculate the first image I using the gray-scale centroid method. k Subpixel coordinates of each LED spot ( ):

[0086] ; ;

[0087] Among them, I mn,i For the first image I k The pixel grayscale value of the i-th LED spot region, m i and n i For the first image I k The column and row numbers of the pixels in the i-th LED spot region, i = 1, 2, 3, 4; the summation range covers the i-th LED spot region.

[0088] Step B, using four sub-pixel coordinates ( ) Calculate the first image I k Geometric center coordinates ( ):

[0089] ;

[0090] .

[0091] Step C, using the geometric center coordinates ( Establish a polar coordinate system with the origin as the origin, and calculate the polar angles of the four LED spots. :

[0092] .

[0093] Step D: Set the four sub-pixel coordinates ( Sort the coordinates in ascending order according to their corresponding polar angles and assign them sequential numbers 1, 2, 3, and 4 to obtain N sets of ordered coordinates.

[0094] With θ k First image I at =0° k For example:

[0095] The sub-pixel coordinates (unordered) of the four extracted LED spots are: (1035.7,778.7), (1526.7,778.7), (1526.7,1268.9), and (1035.7,1268.9).

[0096] Calculated geometric center coordinates ( )for: =1281.2, =1023.8.

[0097] Then calculate the polar angles and assign them numbers: the smallest polar angle is the top left corner, assigned number 1; the second smallest polar angle is the top right corner, assigned number 2; the third smallest polar angle is the bottom right corner, assigned number 3; and the largest polar angle is the bottom left corner, assigned number 4.

[0098] Finally, three sets of ordered coordinates were obtained.

[0099] Step 4: Establish and solve an overdetermined system of equations based on multiple sets of ordered coordinates to obtain the rotation matrix, Euler angles, and translation vector from the collimator world coordinate system to the camera coordinate system when the turntable axis is at 0°, as well as the camera's normalized focal length and principal point coordinates; specifically:

[0100] Based on the above N sets of ordered coordinates, an overdetermined system of equations is established and solved using the Gauss-Newton method or the Levenberg-Marquardt algorithm to obtain the Euler angles of the rotation matrix from the world coordinate system of the collimator to the camera coordinate system when the turntable axis is at 0°. , , Translation vector , , And the camera's x-axis normalized focal length f x The camera's y-axis normalized focal length f y The principal point coordinates of the camera [ , ]:

[0101] ;

[0102] in, The rotation matrix of the collimator 1 relative to the camera 3 when the turntable axis is at 0°; Let n be the translation vector of collimator 1 relative to camera 3 when the turntable axis is at 0°; n is the unit vector of the axis direction. For rotation angle θ k The corresponding rotation matrix; I is a 3×3 identity matrix. T stands for transpose; Let n be an antisymmetric matrix; and These are the rotation matrix and translation vector of the collimator 1 relative to the camera 3 when rotated to the kth time; Let i be the coordinates of the i-th LED in the world coordinate system. , , , ; q=(q x ,q y ,q z ) T f is the point on the axis of rotation closest to the origin of the world coordinate system. x f y These are the normalized focal lengths along the x-axis and y-axis of camera 3, respectively; , [] represents the principal point coordinates of camera 3; the subscripts x, y, and z represent the components of the column vector on the corresponding axes of the camera coordinate system.

[0103] The aforementioned collimator world coordinate system is based on the square array (i.e., a square four-LED array) on the collimator. The origin of this world coordinate system is the center point of the square array. Its x-axis is the horizontal direction of the square array, and the positive x-axis direction is to the right facing the square array. The y-axis is the vertical direction of the square array, and the positive y-axis direction is downward facing the square array. The z-axis is the direction of the central axis of the collimator, and the positive z-axis direction is perpendicular to the plane of the square array and points towards the rotation axis.

[0104] In step 4, there are a total of 16 unknowns, as shown in Table 1:

[0105] Table 1

[0106]

[0107] When N≥3, 8N≥16, which is greater than the number of unknowns (16). The optimal solution for the 16 unknown parameters can be obtained by solving the overdetermined system of equations using methods such as the Levenberg-Marquardt (LM) algorithm or the Gauss-Newton method.

[0108] The solution obtained in this embodiment is: =0.05°, =-0.03°, =0.01°, = = = At the same time, f was also obtained. x =2456.3, f y =2452.1, u0=1280.5, v0=1024.2.

[0109] Step 5: Calculate the six-degree-of-freedom deviation vector of the turntable 5 axis using the rotation matrix, Euler angles, and translation vector of the collimator 1 relative to the camera 3 when the turntable 5 axis is at 0°; specifically:

[0110] Euler angles via rotation matrix , , Translation vector , , Calculate the six-degree-of-freedom deviation vector of the five axes of the turntable. :

[0111] ;

[0112] in, ΔX, ΔY, and ΔZ represent the deviations along the x, y, and z axes, respectively; ΔZ is only used as a monitoring value for axial distance and does not participate in alignment; L design In this embodiment, the preset distance between the optical axis of the collimator 1 and the principal point of the camera 3 under ideal conditions is... ; , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively. , , .

[0113] In this embodiment, ΔX = 0.15 mm, ΔY = -0.08 mm, and ΔZ = 0.2 mm.

[0114] Step 6: Calculate the translation and rotation adjustments using the six-degree-of-freedom deviation vector:

[0115] Translation adjustment amount: ;

[0116] Rotation adjustment amount: .

[0117] In this embodiment, the translation adjustment amount is (No adjustment in the z-direction), rotation adjustment amount is .

[0118] Step 7: Adjust the camera pose in 3D based on the translation and rotation adjustments mentioned above.

[0119] In this embodiment, the camera-collimator spatial pose alignment is completed up to step 7. However, to ensure a more accurate alignment, other embodiments also include a step to verify the alignment of the camera-collimator spatial pose, namely step 8, which specifically includes the following steps:

[0120] Step 8.1: Reacquire N frames of the second image of the four-LED array using the method in Step 2, and recalculate the six-degree-of-freedom deviation vector using the methods in Steps 3 to 5; specifically:

[0121] Reacquire N frames of the second image of the four-LED array using the method in step 2, and recalculate the six-degree-of-freedom deviation vector using the methods in steps 3 to 5. :

[0122] ;

[0123] in, , , These are the deviations on the recalculated x, y, and z axes, respectively. It is used only as a monitoring value for axial distance and is not involved in alignment; , , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively.

[0124] In this embodiment, the rotating shaft rotates again at three angles: −3°, 0°, and +3°.

[0125] Step 8.2: Compare the absolute values ​​of each component in the recalculated six-degree-of-freedom deviation vector with the corresponding preset tolerance thresholds. Determine whether the alignment is complete based on the comparison results. If yes, the alignment is considered complete; otherwise, proceed to step 8.3. Specifically:

[0126] The recalculated six-degree-of-freedom deviation vector The absolute values ​​of each component and their corresponding preset tolerance thresholds , , , , Compare the corresponding values ​​to determine if the following conditions are met:

[0127] ;

[0128] ;

[0129] ;

[0130] ;

[0131] ;

[0132] in, This is the preset tolerance threshold for the deviation on the x-axis. This is the preset tolerance threshold for the deviation on the y-axis. This is the preset tolerance threshold for pitch angle deviation. This is the preset tolerance threshold for yaw angle deviation. This is the preset tolerance threshold for roll angle deviation.

[0133] If all of the above conditions are met, the alignment is considered complete.

[0134] If any of the above conditions are not met, proceed to step 8.3.

[0135] The verification results of this embodiment are shown in Table 2:

[0136] Table 2

[0137]

[0138] The residual deviations have not yet fully met the standards, mainly because the mechanical backflip error of the six-degree-of-freedom adjustment stage 4 and the discretization effect of the minimum step resolution prevent single-time compensation from being fully achieved. Step 8.3: Recalculate the translation and rotation adjustments using the recalculated six-degree-of-freedom deviation vector.

[0139] The recalculated translation adjustment is ;

[0140] The recalculated rotation adjustment amount is .

[0141] In this embodiment, if the system determines that convergence has not occurred, it automatically enters the second iteration, that is, recalculates the translation adjustment and rotation adjustment:

[0142] The recalculated translation adjustment is [-0.035mm +0.025mm 0]. T ;

[0143] The recalculated rotation adjustment is [-0.018° +0.012° -0.003°]. T .

[0144] Step 8.4: Adjust the pose of camera 3 based on the recalculated translation and rotation adjustments, and return to step 8.1 until alignment is determined to be complete.

[0145] In this embodiment, after adjusting the pose of camera 3 based on the recalculated translation and rotation adjustments, a verification calculation is performed again, and the calculation results are shown in Table 3:

[0146] Table 3

[0147]

[0148] Once all deviations are less than the preset tolerance threshold, the system determines that alignment is complete, outputs a "alignment complete" signal, and terminates the closed-loop iteration. Actual tests show that for typical initial deviations (±2mm, ±1°), this method usually converges to the preset tolerance threshold range after 2-3 closed-loop iterations, with each iteration taking no more than 5 seconds.

[0149] In summary, this invention transforms camera-collimator alignment from an open-loop operation relying on human experience into a closed-loop servo alignment process that is automatically controlled by a computer, with quantifiable accuracy and verifiable results. This process is significantly superior to existing technologies in terms of accuracy, efficiency, and repeatability.

[0150] Traditional methods (such as CN103528559A) involve "open-loop adjustment – ​​alignment is considered complete once adjustment is finished," which cannot verify whether the alignment result truly meets the requirements. The embodiments of this invention upgrade alignment from "open-loop operation" to "iterative convergence process" through a closed-loop iteration mechanism, ensuring the reliability of the alignment result.

[0151] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present invention should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A camera-collimator spatial pose alignment method based on a four-LED array and a turntable, characterized in that, Includes the following steps: Step 1: Set up a square four-LED array at the light outlet of the collimator, with the array center on the optical axis of the collimator and the array surface perpendicular to the optical axis. Set the camera on a six-degree-of-freedom adjustment stage and place the six-degree-of-freedom adjustment stage on a turntable. Adjust the six-degree-of-freedom adjustment stage so that all four LED arrays are in the camera's field of view. Step 2: Control the turntable to rotate multiple times. After each rotation, the camera captures one frame of the first image of the four-LED array, resulting in multiple frames of the first image. Step 3: Calculate the sub-pixel coordinates and polar angles of the four LEDs in the first image of each frame, and assign ordered numbers to the four sub-pixel coordinates according to the size of the polar angles to obtain multiple sets of ordered coordinates; Step 4: Establish and solve an overdetermined system of equations based on multiple sets of ordered coordinates to obtain the rotation matrix, Euler angles and translation vector from the world coordinate system of the collimator to the camera coordinate system when the turntable axis is at 0°. Step 5: Calculate the six-degree-of-freedom deviation vector of the turntable axis using the Euler angles of the rotation matrix of the collimator relative to the camera and the translation vector when the turntable axis is at 0°. Step 6: Calculate the translation and rotation adjustments using the six-degree-of-freedom deviation vector; Step 7: Adjust the camera pose based on the translation and rotation adjustments to complete the camera-parallel light tube spatial pose alignment.

2. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 1, characterized in that: Step 1 specifically involves setting up a square four-LED array on a collimator; the side length of the square is 2a, a>0, and the center of the square is located on the optical axis of the collimator, with the horizontal direction being horizontal and the vertical direction being vertical; the camera is set on a six-degree-of-freedom adjustment stage, and the six-degree-of-freedom adjustment stage is set on a turntable; a frame of the four-LED array image is acquired through the camera; the six-degree-of-freedom adjustment stage is adjusted according to the array image so that the four LEDs are located in the four quadrant regions of the array image, and then the entire four-LED array enters the camera's field of view.

3. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 2, characterized in that: Step 2 specifically involves controlling the turntable to rotate N times, where N ≥ 3, with each rotation angle being θ. k , k∈[1,N], θ max / 10≥θ k >0°, θ max The camera's horizontal maximum field of view is defined by the first image I of the four-LED array acquired after each rotation. k , obtain N frames of the first image I k .

4. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 3, characterized in that: Step 3 specifically involves processing the first image I in each frame. k Perform the following processing to obtain N sets of ordered coordinates: Step A: Calculate the first image I using the gray-scale centroid method. k Subpixel coordinates of each LED spot ( ): ； ； Among them, I mn,i For the first image I k The pixel grayscale value of the i-th LED spot region, m i and n i For the first image I k The column and row numbers of the pixels in the i-th LED spot region, i = 1, 2, 3, 4; the summation range covers the i-th LED spot region; Step B, using four sub-pixel coordinates ( ) Calculate the first image I k Geometric center coordinates ( ): ； ； Step C, using the geometric center coordinates ( Establish a polar coordinate system with the origin as the origin, and calculate the polar angles of the four LED spots. : ； Step D: Set the four sub-pixel coordinates ( Sort the coordinates in ascending order according to their corresponding polar angles and assign them sequential numbers 1, 2, 3, and 4 to obtain N sets of ordered coordinates.

5. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 4, characterized in that: Step 4 specifically involves establishing an overdetermined system of equations based on the N sets of ordered coordinates, and solving it using the Gauss-Newton method or the Levenberg-Marquardt algorithm to obtain the Euler angles of the rotation matrix from the collimator world coordinate system to the camera coordinate system when the turntable axis is at 0°. , , Translation vector , , : ； in, The rotation matrix of the collimator relative to the camera when the turntable axis is at 0°; is the translation vector of the collimator relative to the camera when the turntable axis is at 0°; n is the unit vector of the axis direction. For rotation angle θ k The corresponding rotation matrix; I is a 3×3 identity matrix; T is the transpose; Let n be an antisymmetric matrix; and These are the rotation matrix and translation vector of the collimator relative to the camera when it is rotated to the kth time, respectively. Let i be the coordinates of the i-th LED in the world coordinate system. , , , ; q=(q x ,q y ,q z ) T f is the point on the axis of rotation closest to the origin of the world coordinate system. x f y These are the camera's normalized focal length along the x-axis and normalized focal length along the y-axis, respectively; , [] represents the principal point coordinates of the camera; the subscripts x, y, and z represent the components of the column vector on the corresponding axes of the camera coordinate system.

6. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 5, characterized in that: Step 5 specifically involves using the Euler angle rotation matrix. , , Translation vector , , Calculate the six-degree-of-freedom deviation vector of the turntable axis. : ； in, ΔX, ΔY, and ΔZ represent the deviations along the x, y, and z axes of the camera coordinate system, respectively; ΔZ is only used as a monitoring value for axial distance and does not participate in alignment; L design This is a preset distance between the optical axis of the collimator and the principal point of the camera under ideal conditions. , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively. , , .

7. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 6, characterized in that: Step 6 specifically involves using the six-degree-of-freedom deviation vector. Calculate the translation and rotation adjustments: Translation adjustment amount: ; Rotation adjustment amount: .

8. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to any one of claims 1-7, characterized in that, Also includes: Step 8: Verify that the camera-collimator spatial pose is aligned; Step 8.1: Reacquire N frames of the second image of the four-LED array using the method in Step 2, and recalculate the six-degree-of-freedom deviation vector using the methods in Steps 3 to 5. Step 8.2: Compare the absolute values ​​of each component in the recalculated six-degree-of-freedom deviation vector with the corresponding preset tolerance threshold. Determine whether the alignment is complete based on the comparison results. If yes, the alignment is considered complete; otherwise, proceed to step 8.

3. Step 8.3: Recalculate the translation and rotation adjustments using the recalculated six-degree-of-freedom deviation vector; Step 8.4: Adjust the camera pose based on the recalculated translation and rotation adjustments, and return to step 8.1 until alignment is determined to be complete.

9. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 8, characterized in that: Step 8.1 specifically involves re-acquiring N frames of the second image of the four-LED array using the method in Step 2, and recalculating the six-degree-of-freedom deviation vector using the methods in Steps 3 to 5. : ； in, , , These are the deviations on the recalculated x, y, and z axes, respectively. It is used only as a monitoring value for axial distance and is not involved in alignment; , , These are pitch angle deviation, yaw angle deviation, and roll angle deviation, respectively.

10. The camera-collimator spatial pose alignment method based on a four-LED array and a turntable according to claim 9, characterized in that, Also includes: Step 8.2 specifically involves recalculating the six-degree-of-freedom deviation vector. The absolute values ​​of each component and their corresponding preset tolerance thresholds , , , , Compare and determine if the following conditions are met: ； ； ； ； ； in, This is the preset tolerance threshold for the deviation on the x-axis. This is the preset tolerance threshold for the deviation on the y-axis. This is the preset tolerance threshold for pitch angle deviation. This is the preset tolerance threshold for yaw angle deviation. This is the preset tolerance threshold for roll angle deviation; If all of the above conditions are met, the alignment is considered complete. If any of the above conditions are not met, proceed to step 8.3; Step 8.3 specifically involves recalculating the six-degree-of-freedom deviation vector. Recalculate the translation and rotation adjustments.

Images