A multi-camera image taking consistency calibration method, system, device and medium for display screen module manufacturing

Abstract

The application relates to a multi-camera image taking consistency calibration method, system, equipment and medium for display screen module manufacturing, which comprises the following steps: collecting reference images of a calibration board by multiple cameras respectively, performing coarse matching analysis, and reconstructing coarse matching parameters if the matching condition is not met; after the matching is passed, driving a target camera to move and collecting a displacement image sequence, and obtaining coordinate mapping parameters according to the spatial transformation relationship of the displacement image sequence; implementing arbitrary pose disturbance on the target camera and collecting disturbance images, obtaining pose compensation through affine transformation parameter analysis of the reference images and the disturbance images, and combining the coordinate mapping parameters to obtain the pose compensation; and driving the target camera to perform pose compensation operation based on the pose compensation. Through the technical combination of error source isolation correction, multi-degree-of-freedom collaborative calibration and closed-loop reliable verification, the balance problem among precision, efficiency and reliability of the existing calibration method is solved, and technical support is provided for foreign matter removal in display screen module manufacturing.

Description

Technical Field

[0001] This invention relates to the field of camera calibration technology, and more particularly to a multi-camera calibration method for display module manufacturing. Figure One Consistency calibration methods, systems, equipment, and media. Background Technology

[0002] In the manufacturing process of mobile terminal display modules, a multi-layer composite structure is typically laminated to protect the liquid crystal display screen. Conventional processes use a laminated structure of cover glass (CG) and optically clear adhesive (OCA). In this process, if... Figure 1 As shown, the semi-finished product formed by the "soft-on-hard" process consists of a four-layer composite structure from the outside in: a protective film layer, a CG layer, an OCA layer, and a release film layer. Due to particulate contamination in the manufacturing environment, process foreign objects generated during the OCA lamination process will directly lead to display abnormalities. Traditional manual inspection methods suffer from technical bottlenecks such as low inspection efficiency, high false negative rate, and excessive labor costs. Therefore, the adoption of machine vision inspection technology has become an inevitable choice for industry development.

[0003] Existing technologies generally employ dual-camera imaging systems for foreign object detection and localization, acquiring foreign object information from different layers of a composite structure through two imaging devices. The core of this technology lies in requiring strict spatial consistency between the two imaging systems, leading to a complex calibration methodology. A typical industry approach, such as the solution disclosed in patent CN113945575A, involves configuring an XYZ motion mechanism to support the camera group, performing a composite calibration operation of multi-axis translation and planar rotation: first, the optical axis parallelism of the upper and lower cameras is calibrated; then, coordinate system matching and correction are completed through multiple displacements and multi-point image acquisition; finally, a visual parameter correlation model between the two cameras is established.

[0004] However, existing calibration techniques have significant limitations in engineering implementation: First, such as Figure 2 As shown, when there is an axial spacing deviation between the upper and lower cameras and the calibration plate, the geometric measurement based on the pinhole imaging model will produce a nonlinear error, causing the calibration parameters to exceed the tolerance range; secondly, as Figure 3 As shown, the non-orthogonal state of the calibration plate and the camera optical axis, or the non-parallel configuration of the optical axes of the two cameras, will cause perspective projection distortion, severely degrading the calibration accuracy. More significantly, the multi-shift image acquisition strategy used in traditional methods to achieve calibration accuracy not only significantly prolongs the calibration cycle but also degrades system stability due to the cumulative effect of errors. These technical defects directly restrict the deployment efficiency and operational reliability of the detection system, urgently requiring innovative solutions. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides a multi-camera consistency calibration method for display module manufacturing Figure One The consistency calibration method, system, device and medium solve the technical problems that the existing double-camera calibration method causes calibration precision deterioration due to interval deviation and non-orthogonal optical axis, error accumulation and low efficiency caused by multiple superimposed shifts, and is difficult to meet the high-efficiency and high-precision detection requirements of display modules.

[0007] (II) Technical solutions

[0008] To achieve the above-mentioned purposes, the main technical solutions adopted by the present application include:

[0009] In a first aspect, the embodiments of the present application provide a multi-camera consistency calibration method for display module manufacturing Figure One The consistency calibration method comprises:

[0010] The reference images of the calibration board fixed on the detection platform are respectively collected by the multi-camera, coarse matching analysis based on geometric feature distance is performed on the reference images, and if the preset matching condition is not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters;

[0011] After the coarse matching passes, the target camera in the multi-camera is driven to perform a motion of a predetermined displacement amount along the orthogonal direction and collect a displacement image sequence for the display module, and according to the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from the physical space to the image space are solved;

[0012] Any pose disturbance is implemented on the target camera and a disturbance image is collected, the pose compensation amount containing the displacement compensation amount and the rotation correction angle is obtained by analyzing the affine transformation parameters of the reference image and the disturbance image in combination with the coordinate mapping parameters;

[0013] Based on the pose compensation amount, the target camera is driven to perform a pose compensation operation containing reverse displacement and rotation, a verification image set is collected in the imaging space of the compensated multi-camera, and the multi-camera consistency calibration parameters are generated according to the geometric feature alignment degree of the verification image set.

[0014] Optionally, the reference images of the calibration board fixed on the detection platform are respectively collected by the multi-camera, coarse matching analysis based on geometric feature distance is performed on the reference images, and if the preset matching condition is not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters, which comprises:

[0015] The first reference image and the second reference image of the calibration board fixed on the detection platform are respectively collected by the upper camera and the lower camera;

[0016] Sub-pixel coordinates of feature points in each reference image are extracted respectively to generate candidate point pairs under the view of multi-camera;

[0017] A symmetric distance matrix between reference images is obtained by calculating the Euclidean distance of each pair of feature points, and each row of distance matrix elements in the symmetric distance matrix is compared for cross-camera similarity;

[0018] When each row of distance matrix elements in the symmetric distance matrix between reference images meets the preset coarse matching condition, it is marked as a coarse matching point pair;

[0019] When each row of distance matrix elements in the symmetric distance matrix between reference images does not meet the preset coarse matching condition, the distance tolerance threshold is adjusted dynamically or the calibration plate / camera plane is rotated by a linkage mechanism to reconstruct the coarse matching parameters until the preset coarse matching condition is met;

[0020] wherein,

[0021] The upper camera is detachably mounted at the end of the driving mechanism, and the driving mechanism realizes the adjustment of the degrees of freedom of at least two-dimensional plane translation and rotation around the optical axis, and the optical axis points to the imaging device on the upper surface of the detected object;

[0022] The lower camera is fixedly installed below the detection platform, and the optical axis is perpendicular to the platform bearing surface and points to the imaging device on the lower surface of the detected object;

[0023] The calibration plate is configured with an array of geometric features, the distance between each feature point is known, and the arrangement pattern has local uniqueness;

[0024] The coarse matching condition is that the distance elements in a row of a symmetric distance matrix and the corresponding distance elements in a row of another symmetric distance matrix are compared across the matrix rows, and when there are at least three elements with a difference less than a preset threshold, the feature point pairs corresponding to the two rows are determined as coarse matching point pairs.

[0025] Optionally, after the matching is passed, after the coarse matching is passed, the target camera in the multi-camera is driven to move in the orthogonal direction by a predetermined displacement and a displacement image sequence for the display screen module is collected, and the coordinate mapping parameters from the physical space to the image space are obtained according to the spatial transformation relationship of the displacement image sequence in different orthogonal directions, including:

[0026] The upper camera in the multi-camera is moved by a first predetermined physical displacement in the first orthogonal direction by the preset driving mechanism, a first orthogonal displacement image is collected, a first matching matrix containing displacement components is analyzed based on the matching point set of the second reference image set and the first orthogonal displacement image by using a robust estimation algorithm;

[0027] The camera is controlled by a preset driving mechanism to move along a second orthogonal direction by a predetermined displacement and collect a second displacement image; a robust estimation algorithm is used to analyze a second matching matrix containing displacement components based on a matching point set of the first displacement image and the second displacement image;

[0028] A spatial constraint equation of the first orthogonal direction is established according to a mapping relationship between the first predetermined physical displacement and the first matching matrix;

[0029] A spatial constraint equation of the second orthogonal direction is established according to a mapping relationship between the second predetermined physical displacement and the second matching matrix;

[0030] The coordinate mapping parameters from the physical space to the image space are calculated by simultaneously solving the spatial constraint equation of the first orthogonal direction and the spatial constraint equation of the second orthogonal direction.

[0031] Optionally, the first matching matrix is:

[0032]

[0033] In the formula, H1 is the first matching matrix, and components represent that a pixel displacement of Δu1 in size is generated in the horizontal direction of the image coordinate system, and a pixel displacement of Δv1 in size is generated in the vertical direction of the image coordinate system;

[0034] The spatial constraint equation of the first orthogonal direction is:

[0035]

[0036] In the formula, represent that the displacement of the camera in the first orthogonal direction of the physical coordinate system is a, the displacement in the second orthogonal direction of the physical coordinate system is 0, and T 00 is a mapping proportion coefficient of the physical displacement in the first orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system, T 10 is a mapping proportion coefficient of the physical displacement in the first orthogonal direction of the physical coordinate system to the vertical direction of the image coordinate system, T 01 is a mapping proportion coefficient of the physical displacement in the second orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system, and T 11 is a mapping proportion coefficient of the physical displacement in the second orthogonal direction of the physical coordinate system to the vertical direction of the image coordinate system;

[0037] The second matching matrix is:

[0038]

[0039] In the formula, H2 is the second matching matrix, and components represents that a pixel displacement amount of Δu2 in the horizontal direction of the image coordinate system is generated, and a pixel displacement amount of Δv2 in the vertical direction of the image coordinate system is generated;

[0040] The spatial constraint equation of the second orthogonal direction is:

[0041]

[0042] In the formula, represents that the displacement of the camera in the first orthogonal direction of the physical coordinate system is 0, and the displacement of the camera in the second orthogonal direction of the physical coordinate system is b;

[0043] The translation mapping parameter is:

[0044]

[0045] Optionally, any pose disturbance is implemented on the target camera, and a disturbed image is collected, the pose compensation quantity including the displacement compensation quantity and the rotation correction angle is obtained by analyzing the affine transformation parameter of the reference image and the disturbed image, and combining the coordinate mapping parameter.

[0046] The preset driving mechanism controls the upper camera to generate any pose disturbance, and a disturbed image after the pose disturbance is collected;

[0047] Based on the matching point set of the disturbed image and the first reference image, a robust estimation algorithm is used to analyze the third matching matrix including the displacement component and the rotation component;

[0048] According to the translation component of the third matching matrix and the coordinate mapping parameter, the displacement compensation quantity in the physical space is calculated and generated;

[0049] The third matching matrix is normalized by using the camera intrinsic parameter matrix pre-calibrated, and the rotation component of the normalized transformation matrix is decomposed to obtain the rotation correction angle;

[0050] The pose compensation quantity of the multi-camera pose consistency is generated by fusing the displacement compensation quantity and the rotation correction angle.

[0051] Optionally, the third matching matrix is:

[0052]

[0053] In the formula, H3 is the third matching matrix, h 11 is a scaling and rotation contribution component of the horizontal direction displacement of the physical coordinate system to the horizontal direction of the image coordinate system, h 12 is a shearing and rotation contribution component of the vertical direction displacement of the physical coordinate system to the horizontal direction of the image coordinate system, h 21 is a shearing and rotation contribution component of the horizontal direction displacement of the physical coordinate system to the vertical direction of the image coordinate system, h22 h represents the contribution of the displacement in the vertical direction of the physical coordinate system to the scaling and rotation in the vertical direction of the image coordinate system. 13 h represents the horizontal displacement component in the image coordinate system. 23 h represents the vertical displacement component in the image coordinate system. 31 h represents the horizontal projection component of the perspective distortion caused by the tilt of the calibration plate / camera. 32 The vertical projection component of the perspective distortion caused by the tilt of the calibration plate / camera;

[0054] The displacement compensation amount in physical space is:

[0055]

[0056] In the formula, x1 is the displacement that the upper camera needs to compensate for in the horizontal direction of the physical coordinate system, and y1 is the displacement that the upper camera needs to compensate for in the vertical direction of the physical coordinate system.

[0057] Normalized transformation matrix:

[0058] H norm =K -1 *H3*K;

[0059] In the formula, H norm Here, K is the normalized transformation matrix, and K is the pre-calibrated camera intrinsic parameter matrix.

[0060] The rotation correction angle is:

[0061]

[0062] In the formula, C is the rotation correction angle, and H norm [0,0] represents the element in the first row and first column of the normalized matrix, H norm [1,0] represents the element in the second row and first column of the normalized matrix.

[0063] Optionally, the first matching matrix, the second matching matrix, and the third matching matrix are all obtained through the following steps:

[0064] Extract sub-pixel level feature point coordinates from the image pairs to be matched, and generate candidate point pairs from multiple camera perspectives;

[0065] The Euclidean distance between each pair of feature points is calculated to obtain the symmetric distance matrix between the pairs of images to be matched. Cross-camera similarity comparison is then performed on the distance matrix elements in each row of the symmetric distance matrix.

[0066] When each row of the distance matrix element in the symmetric distance matrix satisfies the preset coarse matching condition, it is marked as a coarse matching point pair;

[0067] When the distance matrix elements in each row of the symmetric distance matrix do not meet the preset coarse matching conditions, the coarse matching parameters are reconstructed by dynamically adjusting the distance tolerance threshold or by rotating the calibration plate / camera plane through a linkage mechanical mechanism until the preset coarse matching conditions are met.

[0068] A polar coordinate neighborhood is determined with each candidate point as the center. Neighboring points are selected in the polar coordinate neighborhood and the polar angles of each neighborhood point relative to the candidate points are calculated. After sorting, a polar angle sequence representing the distribution relationship of neighborhood angles across camera views is obtained.

[0069] Perform differential angle transformation on the polar angle sequence across camera views to eliminate rotational bias and generate the obtained differential angle sequence;

[0070] Valid matching point pairs are determined by comparing the similarity of the differential angle sequences. When the similarity of the differential angle sequences exceeds the second threshold, they are determined to be valid matching point pairs. If there are multiple valid matching point pairs, the point pair with the highest similarity in neighborhood geometric structure is selected to form the final matching point set.

[0071] The homography matrix is ​​solved by a combination algorithm of random sampling consistency and least squares method for the final matching point set to obtain the first matching matrix, the second matching matrix, or the third matching matrix.

[0072] Specifically, when calculating the first matching matrix, the images to be matched are the second reference image and the first orthogonal displacement image; when calculating the second matching matrix, the images to be matched are the first displacement image and the second displacement image; and when calculating the third matching matrix, the images to be matched are the perturbation image and the first reference image.

[0073] Secondly, embodiments of the present invention provide a multi-camera capture method for display module manufacturing. Figure One Consistency calibration system, including:

[0074] The coarse matching module is used to acquire reference images of the calibration board fixed on the detection platform by multiple cameras, and to perform coarse matching analysis based on geometric feature distance on the reference images. If the preset matching conditions are not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters.

[0075] The coordinate mapping module is used to drive the target camera in the multi-camera system to perform a predetermined displacement along the orthogonal direction and acquire the displacement image sequence after the coarse matching is passed. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from physical space to image space are obtained.

[0076] The pose compensation module is used to perform arbitrary pose perturbation on the target camera and acquire the perturbation image. By analyzing the affine transformation parameters of the reference image and the perturbation image and combining the coordinate mapping parameters, the pose compensation amount, which includes the displacement compensation amount and the rotation correction angle, is obtained.

[0077] The camera consistency calibration module is used to drive the target camera to perform pose compensation operations including reverse displacement and rotation based on the pose compensation amount. It acquires a verification image set through the compensated multi-camera imaging space and generates multi-camera consistency calibration parameters based on the geometric feature alignment of the verification image set.

[0078] Thirdly, embodiments of the present invention provide a multi-camera acquisition method. Figure One A consistency calibration device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the multi-camera acquisition described above for manufacturing display modules. Figure One Consistency calibration method.

[0079] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the multi-camera acquisition method for display module manufacturing as described above. Figure One Consistency calibration method.

[0080] (III) Beneficial Effects

[0081] The beneficial effects of this invention are: by combining a hierarchical and progressive calibration strategy with a dynamic adaptive feedback mechanism, this invention significantly improves the accuracy and efficiency of multi-camera calibration, specifically manifested in:

[0082] First, based on coarse matching analysis of geometric feature distances, combined with a collaborative mechanism of dynamic threshold adjustment and pose reconstruction, the problem of initial matching inaccuracies caused by assembly deviations or environmental vibrations is effectively addressed, significantly enhancing the stability of parameter initialization under complex working conditions. On this basis, through the coupled analysis of orthogonal displacement driving and spatial transformation relationships, a high-precision mapping model from physical space to image space is quickly established, avoiding redundant operations caused by traditional multi-point image acquisition. Simultaneously, it eliminates nonlinear calibration errors caused by camera spacing deviations and non-orthogonal optical axes, significantly shortening the calibration cycle and suppressing error accumulation effects.

[0083] Next, to address the rotation-translation coupling error introduced by mechanical disturbances, the analytical method of affine transformation parameters and the coordinate mapping model are integrated to achieve synchronous calculation of displacement compensation and rotation correction angle, breaking through the technical bottleneck of error propagation in traditional step-by-step calibration. At the same time, after performing reverse pose compensation, the calibration quality is dynamically evaluated by verifying the geometric feature alignment of the image set, forming a feedback-driven parameter calibration link to ensure the long-term stability of the multi-camera system in continuous operation.

[0084] Therefore, this invention systematically overcomes the challenge of balancing accuracy, efficiency, and reliability in existing calibration methods by combining error source isolation correction, multi-degree-of-freedom collaborative calibration, and closed-loop reliable verification, providing reliable technical support for foreign object removal in display module manufacturing. Attached Figure Description

[0085] Figure 1 This is a diagram illustrating foreign objects generated during the manufacturing process of a display module, as provided in an embodiment of the present invention.

[0086] Figure 2 This is a schematic diagram showing that the upper and lower cameras are not at the same distance from the CG support platform, as provided in an embodiment of the present invention.

[0087] Figure 3 A schematic diagram showing that the calibration plate of the method provided in the embodiments of the present invention is not perpendicular to the viewing angles of the two cameras, or the viewing angles of the two cameras are not parallel.

[0088] Figure 4 This is a schematic diagram of the overall process of the method provided in the embodiments of the present invention;

[0089] Figure 5 This is a schematic diagram illustrating the specific process of step S1 of the method provided in this embodiment of the invention;

[0090] Figure 6 A schematic diagram of a calibration plate for the method provided in an embodiment of the present invention;

[0091] Figure 7 This is a detailed flowchart illustrating step S2 of the method provided in this embodiment of the invention;

[0092] Figure 8 A schematic diagram illustrating the specific process of step S3 of the method provided in this embodiment of the invention. Detailed Implementation

[0093] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0094] like Figure 4 As shown, an embodiment of the present invention proposes a multi-camera acquisition method. Figure OneThe consistency calibration method includes: acquiring reference images of a calibration plate fixed to a detection platform using multiple cameras; performing coarse matching analysis based on geometric feature distance on the reference images; dynamically adjusting the matching threshold or adjusting the spatial pose of the rotating calibration plate / camera to reconstruct the coarse matching parameters if the preset matching conditions are not met; after successful coarse matching, driving the target camera in the multiple cameras to perform a predetermined displacement along an orthogonal direction and acquiring a displacement image sequence for the display module; obtaining coordinate mapping parameters from physical space to image space based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions; subjecting the target camera to arbitrary pose perturbation and acquiring perturbation images; obtaining a pose compensation amount including displacement compensation and rotation correction angle by analyzing the affine transformation parameters of the reference image and the perturbation image, combined with the coordinate mapping parameters; driving the target camera to perform pose compensation operations including reverse displacement and rotation based on the pose compensation amount; acquiring a verification image set through the compensated multi-camera imaging space; and generating multi-camera consistency calibration parameters based on the geometric feature alignment of the verification image set.

[0095] This invention significantly improves the accuracy and efficiency of multi-camera calibration by combining a hierarchical, progressive calibration strategy with a dynamic adaptive feedback mechanism. Specifically, this is manifested in:

[0096] First, based on coarse matching analysis of geometric feature distances, combined with a collaborative mechanism of dynamic threshold adjustment and pose reconstruction, the problem of initial matching inaccuracies caused by assembly deviations or environmental vibrations is effectively addressed, significantly enhancing the stability of parameter initialization under complex working conditions. On this basis, through the coupled analysis of orthogonal displacement driving and spatial transformation relationships, a high-precision mapping model from physical space to image space is quickly established, avoiding redundant operations caused by traditional multi-point image acquisition. Simultaneously, it eliminates nonlinear calibration errors caused by camera spacing deviations and non-orthogonal optical axes, significantly shortening the calibration cycle and suppressing error accumulation effects.

[0097] Next, to address the rotation-translation coupling error introduced by mechanical disturbances, the analytical method of affine transformation parameters and the coordinate mapping model are integrated to achieve synchronous calculation of displacement compensation and rotation correction angle, breaking through the technical bottleneck of error propagation in traditional step-by-step calibration. At the same time, after performing reverse pose compensation, the calibration quality is dynamically evaluated by verifying the geometric feature alignment of the image set, forming a feedback-driven parameter calibration link to ensure the long-term stability of the multi-camera system in continuous operation.

[0098] Therefore, this invention systematically overcomes the challenge of balancing accuracy, efficiency, and reliability in existing calibration methods by combining error source isolation correction, multi-degree-of-freedom collaborative calibration, and closed-loop reliable verification, providing reliable technical support for foreign object removal in display module manufacturing.

[0099] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0100] Specifically, embodiments of the present invention provide multi-camera acquisition. Figure One Consistency calibration methods, including:

[0101] S1. Acquire reference images of the calibration plate fixed on the detection platform using multiple cameras. Perform coarse matching analysis based on geometric feature distance on the reference images. If the preset matching conditions are not met, dynamically adjust the matching threshold or adjust the spatial pose of the rotating calibration plate / camera to reconstruct the coarse matching parameters.

[0102] Furthermore, such as Figure 5 As shown, step S1 includes:

[0103] S11. Acquire the first reference image and the second reference image of the calibration plate fixed on the detection platform by using the upper and lower cameras respectively.

[0104] First, it needs to be clarified that the upper camera is a detachable device mounted at the end of the drive mechanism, which enables at least two-dimensional planar translation and rotation around the optical axis, with the optical axis pointing towards the upper surface of the object being inspected; the lower camera is a device fixedly mounted below the inspection platform, with the optical axis perpendicular to the platform's bearing surface and pointing towards the lower surface of the object being inspected.

[0105] refer to Figure 6 It is known that the calibration board is equipped with a geometric feature array, the spacing between each feature point is known, and the arrangement pattern is locally unique. The calibration board is placed on the CG platform, and a first reference image is obtained using the lower camera, and a second reference image is obtained using the upper camera. A coarse matching is performed on the first and second reference images. If the coarse matching fails, it indicates that the positions of the two cameras relative to the calibration board are not necessarily aligned. If increasing the coarse matching threshold still fails to achieve a match, it indicates that the calibration board plane or the camera plane is not horizontal. If the camera or calibration board is rotated, and then a distance matrix matching is performed on the images before and after rotation, if the match fails, the corresponding device (camera or calibration board) is not horizontal.

[0106] S12. Extract the sub-pixel level coordinates of feature points in each reference image to generate candidate point pairs from multiple camera perspectives.

[0107] S13. Calculate the Euclidean distance between each pair of feature points to obtain the symmetric distance matrix between the reference images, and perform cross-camera similarity comparison on each row of the distance matrix elements in the symmetric distance matrix.

[0108] S14. When each row of distance matrix elements in the symmetric distance matrix between reference images satisfies the preset coarse matching condition, it is marked as a coarse matching point pair.

[0109] S15. When the distance matrix elements in each row of the symmetric distance matrix between the reference images do not meet the preset coarse matching conditions, the coarse matching parameters are reconstructed by dynamically adjusting the distance tolerance threshold or by rotating the calibration plate / camera plane using a linkage mechanical mechanism until the preset coarse matching conditions are met. The coarse matching conditions are as follows: Cross-matrix row comparisons are performed in the symmetric distance matrix between the reference images. If at least three elements in the set of distance elements in a row of one symmetric distance matrix and the corresponding set of distance elements in a row of another symmetric distance matrix have a difference less than a preset threshold, the feature point pair corresponding to those two rows is determined to be a coarse matching point pair.

[0110] In one embodiment, the coarse matching process includes: first, detecting the center coordinates (x, y) of circles in each image. i ,y i The accuracy is improved through sub-pixel optimization, generating a feature point set containing sub-pixel level coordinates; subsequently, a point set S = {p1, p2, ..., p...} is generated for each viewpoint. m}, where p i =(x i ,y i ), calculate the Euclidean distance between each pair of feature points. First, identify the center of the marker point on the calibration board for the upper and lower camera images, and calculate the spatial distance between any two points respectively, constructing symmetric matrices A (distance matrix of the first image) and B (distance matrix of the second image).

[0111] Next, by comparing the row vector similarity of the distance matrices from the perspectives of the two cameras, candidate matching point pairs that meet the preset distance difference threshold are selected. Specifically, the coarse matching condition is as follows: set a threshold t1, compare the distance elements of the i-th row in matrix A with each row in matrix B. If more than 3 (including 3) elements in the i-th row of matrix A have a difference less than t1 with the j-th row of matrix B, then S is considered to be matched. i With T j For candidate pairs with the same name.

[0112] Points are identified as candidate matching pairs. If the number of candidate point pairs is insufficient, the matching parameters are reconstructed until the conditions are met by dynamically adjusting the distance tolerance threshold or mechanically correcting the calibration plate / camera tilt angle. This process forms a stable coarse matching basis through quantitative comparison of the distance matrix and dynamic threshold adjustment.

[0113] S2. After the coarse matching is successful, drive the target camera in the multi-camera system to perform a predetermined displacement along the orthogonal direction and acquire a displacement image sequence for the display module. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, obtain the coordinate mapping parameters from physical space to image space.

[0114] Furthermore, such as Figure 7 As shown, step S2 includes:

[0115] S21. Control the upper camera in the multi-camera system to move a first predetermined physical displacement along the first orthogonal direction through a preset drive mechanism, acquire the first orthogonal displacement image, and based on the matching point set of the second reference image set and the first orthogonal displacement image, use a robust estimation algorithm to parse the first matching matrix containing the displacement component.

[0116] S22. Control the upper camera to move a predetermined displacement along the second orthogonal direction and acquire the second displacement image through a preset drive mechanism. Based on the matching point set of the first displacement image and the second displacement image, use a robust estimation algorithm to parse the second matching matrix containing the displacement component.

[0117] S23. Based on the mapping relationship between the first predetermined physical displacement and the first matching matrix, establish the spatial constraint equation for the first orthogonal direction.

[0118] S24. Based on the mapping relationship between the second predetermined physical displacement and the second matching matrix, establish the spatial constraint equation for the second orthogonal direction.

[0119] S25. Simultaneously solve the spatial constraint equations of the first orthogonal direction and the spatial constraint equations of the second orthogonal direction to calculate the coordinate mapping parameters from physical space to image space.

[0120] In one specific embodiment, if the coarse matching is successful, the motion mechanism is manipulated to move the upper camera a distance 'a' along the positive x-direction, capturing a first orthogonal displacement image. A set of matching points between the second reference image set and the first orthogonal displacement image is obtained through both coarse and fine matching. Then, a robust estimation algorithm is used to analyze and obtain the first matching matrix H1 between the second reference image set and the first orthogonal displacement image. Since there is no rotation, the form of the matching matrix H1 on the image is as follows:

[0121]

[0122] In the formula, H1 is the first matching matrix, and the components are... This indicates a pixel displacement of magnitude Δu1 in the horizontal direction of the image coordinate system and a pixel displacement of magnitude Δv1 in the vertical direction of the image coordinate system, both in pixels.

[0123] The camera's displacement in the physical coordinate system is Then the spatial constraint equation for the first orthogonal direction is:

[0124]

[0125] In the formula, Let T represent the displacement of the camera in the first orthogonal direction of the physical coordinate system as 'a', and the displacement in the second orthogonal direction of the physical coordinate system as 0. 00 T is the scaling factor for the physical displacement in the first orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system. 10 T is the scaling factor for the physical displacement in the first orthogonal direction of the physical coordinate system to the perpendicular direction of the image coordinate system. 01 T is the scaling factor for the physical displacement in the second orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system. 11 It is the mapping scaling factor of the physical displacement in the second orthogonal direction of the physical coordinate system to the vertical direction of the image coordinate system.

[0126] Then, the motion mechanism is manipulated to move the upper camera a distance b along the positive y-direction of the motion mechanism, and a second orthogonal displacement image is obtained. Through coarse and fine matching, the matching point set between the second reference image set and the first orthogonal displacement image is obtained. Next, a robust estimation algorithm is used to analytically obtain the second matching matrix H2 between the first and second displacement images. Since there is no rotation, the form of the matching matrix H2 on the image is as follows:

[0127]

[0128] In the formula, H2 is the second matching matrix, and the components are... This indicates that a pixel displacement of magnitude Δu2 is generated in the horizontal direction of the image coordinate system, and a pixel displacement of magnitude Δv2 is generated in the vertical direction of the image coordinate system.

[0129] The camera's displacement in the physical coordinate system is Then the spatial constraint equation for the second orthogonal direction is:

[0130]

[0131] In the formula, This indicates that the camera's displacement in the first orthogonal direction of the physical coordinate system is 0, and its displacement in the second orthogonal direction of the physical coordinate system is b.

[0132] The expression for the coordinate mapping parameter T from physical coordinates to image coordinates, obtained by simultaneously solving equations 1 and 2, is as follows:

[0133]

[0134] S3. Perform arbitrary pose perturbation on the target camera and acquire the perturbation image. Analyze the affine transformation parameters of the reference image and the perturbation image, and combine them with the coordinate mapping parameters to obtain the pose compensation amount, which includes the displacement compensation amount and the rotation correction angle.

[0135] Furthermore, such as Figure 8 As shown, step S3 includes:

[0136] S31. Control the upper camera to generate arbitrary pose disturbance through a preset drive mechanism, and acquire the disturbance image after pose disturbance.

[0137] S32. Based on the matching point set between the perturbed image and the first reference image, a robust estimation algorithm is used to parse the third matching matrix containing displacement and rotation components.

[0138] S33. Calculate the displacement compensation amount in the physical space based on the translation components and coordinate mapping parameters of the third matching matrix.

[0139] S34. Normalize the third matching matrix using the pre-calibrated camera intrinsic parameter matrix, and then decompose the normalized transformation matrix into rotational components to obtain the rotation correction angle.

[0140] S35. Fuse displacement compensation and rotation correction angle to generate pose compensation for multi-camera pose consistency.

[0141] In a specific embodiment, the upper camera is moved to an arbitrary initial position to take a picture and obtain a perturbed image. The third matching matrix H3 between the first reference image and the perturbed image is calculated. H3 may contain both translation and rotation. Assuming that the alignment coordinates of corresponding points in the upper and lower camera images are [x',y',1]T and [x,y,1]T, the point pairs on these two images from different perspectives can be represented by a projective transformation H3, that is: [x',y',1]T = H*[x,y,1]T. The expression of the third matching matrix H3 is as follows:

[0142]

[0143] In the formula, H3 is the third matching matrix, h 11 h represents the contribution of the horizontal displacement of the physical coordinate system to the scaling and rotation of the image coordinate system in the horizontal direction. 12 h represents the contribution of the vertical displacement of the physical coordinate system to the shearing and rotation of the horizontal direction of the image coordinate system. 21 h represents the contribution of the horizontal displacement of the physical coordinate system to the shearing and rotation in the vertical direction of the image coordinate system. 22 h represents the contribution of the displacement in the vertical direction of the physical coordinate system to the scaling and rotation in the vertical direction of the image coordinate system. 13 h represents the horizontal displacement component in the image coordinate system.23 h represents the vertical displacement component in the image coordinate system. 31 h represents the horizontal projection component of the perspective distortion caused by the tilt of the calibration plate / camera. 32 The vertical projection component of the perspective distortion caused by the tilt of the calibration plate / camera.

[0144] The displacement compensation amount in physical space is:

[0145]

[0146] In the formula, x1 is the displacement that the upper camera needs to compensate in the horizontal direction of the physical coordinate system, which is used to eliminate the horizontal imaging offset caused by initial pose deviation or mechanical error, and y1 is the displacement that the upper camera needs to compensate in the vertical direction of the physical coordinate system, which is used to correct the imaging misalignment in the vertical direction.

[0147] Normalized homography matrix: Given the camera intrinsic parameter matrix K, calculate the normalized homography matrix:

[0148] H norm =K -1 *H3*K;

[0149] Then, at this time H norm Including only extrinsic parameters and scene plane parameters, the geometric meaning is purer, facilitating subsequent decomposition to obtain camera pose (such as translation direction and rotation angle). Therefore, from H... norm Extract the top-left 2x2 submatrix:

[0150] The expression for the rotation angle θ at this time is: In the formula, C is the rotation correction angle, and H norm [0,0] represents the element in the first row and first column of the normalized matrix, H norm [1,0] represents the element in the second row and first column of the normalized matrix.

[0151] It is important to understand that the first matching matrix, the second matching matrix, and the third matching matrix are all obtained through the following steps:

[0152] A11. Extract sub-pixel level feature point coordinates from the image pairs to be matched, and generate candidate point pairs from multiple camera perspectives. Specifically, when calculating the first matching matrix, the images to be matched are the second reference image and the first orthogonal displacement image; when calculating the second matching matrix, the images to be matched are the first displacement image and the second displacement image; and when calculating the third matching matrix, the images to be matched are the perturbed image and the first reference image.

[0153] A12. Calculate the Euclidean distance between each pair of feature points to obtain the symmetric distance matrix between the pairs of images to be matched, and perform cross-camera similarity comparison on each row of the distance matrix elements in the symmetric distance matrix.

[0154] A13. When each row of distance matrix elements in the symmetric distance matrix satisfies the preset coarse matching condition, it is marked as a coarse matching point pair.

[0155] A13. When the distance matrix elements in each row of the symmetric distance matrix do not meet the preset coarse matching conditions, the coarse matching parameters are reconstructed by dynamically adjusting the distance tolerance threshold or by rotating the calibration plate / camera plane through a linkage mechanical mechanism until the preset coarse matching conditions are met.

[0156] A14. Determine the polar coordinate neighborhood with each candidate point as the center, select neighborhood points in the polar coordinate neighborhood and calculate the polar angle of each neighborhood point relative to the candidate point. After sorting the angle values ​​from smallest to largest, obtain the polar angle sequence across camera view that represents the distribution relationship of neighborhood angles.

[0157] A15. Perform differential angle transformation on the polar angle sequence across camera viewpoints to eliminate rotational bias and generate the obtained differential angle sequence.

[0158] A16. Valid matching point pairs are determined by comparing the similarity of the differential angle sequences. When the similarity of the differential angle sequences exceeds the second threshold, they are determined to be valid matching point pairs. If there are multiple valid matching point pairs, the point pair with the highest similarity in the neighborhood geometric structure is selected to form the final matching point set.

[0159] A17. When the similarity does not exceed the second threshold, a secondary adjustment of the calibration board or camera pose is triggered and the coarse matching analysis is re-executed.

[0160] A18. Solve the homography matrix using a combination algorithm of random sampling consistency and least squares for the final matching point set to obtain the first matching matrix, the second matching matrix, or the third matching matrix.

[0161] In another specific example, the first matching matrix, the second matching matrix, and the third matching matrix are all obtained through the following steps: First, the center coordinates (x, y) of each circle in each image (including the reference image, the displacement image, and the perturbed image) are detected. i ,y i To improve accuracy, sub-pixel optimization is used to generate a point set S = {p1, p2, ..., p...} for each viewpoint. m}, where p i =(x i ,y i ).

[0162] Calculate the Euclidean distance between each pair of points. Construct symmetric matrices A (distance matrix of the first image) and B (distance matrix of the second image).

[0163] Next, coarse matching is performed. For each row of matrices A and B (corresponding to the neighborhood distance set of a single feature point), cross-camera similarity comparison is performed. A first threshold t1 is set. If the distance difference between two rows exceeds 3 times ≤ t1, they are marked as candidate matching point pairs (S). i ,T j If the coarse matching condition is not met, the distance tolerance threshold is dynamically expanded or the mechanical mechanism is driven to rotate the calibration plate / camera plane to reconstruct the matching parameters until a sufficient number of candidate matching point pairs are obtained.

[0164] Then, taking each candidate point as the center, select feature points p in the neighborhood. k Calculate its relative to the center point p i polar angle Where θ i,k For the neighborhood point p i Relative to the central reference point p k The polar angle (azimuth), (x k ,y k Let p be a point. k The coordinates, (x i ,y i Let p be a point. i The coordinates.

[0165] Construct an angle matrix A representing the distribution relationship of neighborhood angles. θ and B θ Store the angular relationships of the neighborhood of each point.

[0166] Furthermore, the fine matching process begins: matching the neighborhood angle matrices of candidate point pairs is performed, and the difference angle is calculated after sorting the polar angle sequences (eliminating the influence of rotation); if the similarity of the difference angle sequences exceeds a threshold (e.g., cosine similarity > 0.95), it is determined to be a correct match; in the case of multiple candidate matches, the optimal matching point pair is selected based on the similarity of the neighborhood geometric structure (consistency between distance and angle distribution) to form the final matching point set.

[0167] Finally, for the final matching point set {(S i ,T j The RANSAC (Random Sample Consistency) algorithm is used to remove outliers, and the homography matrix H is calculated using the least squares method.

[0168]

[0169] Therefore, this embodiment effectively solves the matching ambiguity and noise interference problems in multi-layer composite structure imaging through a three-level progressive mechanism: coarse matching dynamic adjustment → fine matching angle verification → robust homography matrix solution. Specifically, sub-pixel optimization and neighborhood geometric constraints ensure feature point localization accuracy and matching reliability; differential angle transformation eliminates the influence of camera rotation deviation on matching; and the RANSAC + least squares combined algorithm balances outlier robustness and parameter optimization, providing a high-precision transformation model for multi-camera spatial alignment.

[0170] S4. Based on the pose compensation amount, drive the target camera to perform pose compensation operations including reverse displacement and rotation, acquire a verification image set through the compensated multi-camera imaging space, and generate multi-camera consistency calibration parameters based on the geometric feature alignment of the verification image set.

[0171] After obtaining the pose compensation values ​​including displacement compensation amounts x1 and y1 and rotation correction angle θ, the following operations are performed to achieve consistent calibration of the multi-camera imaging space:

[0172] The control motion mechanism drives the upper camera to move backward along the X-axis of the coordinate system by |x1| and backward along the Y-axis by |y1|, eliminating the imaging translation error caused by the initial pose deviation. Simultaneously, the upper camera is controlled to rotate clockwise around the optical axis by |θ| to correct the image distortion caused by the rotation of the viewing angle.

[0173] After the pose compensation operation is completed, the upper camera acquires a set of compensated verification images, and the lower camera is simultaneously triggered to acquire a set of reference images of the same calibration board. Subpixel-level feature point extraction (such as circle center coordinates) is performed on the verification image set and the reference image set to generate a set of matching point pairs. The coordinate residuals of all matching point pairs are calculated, and the coordinate residual statistics (mean and standard deviation) of the matching point pairs are calculated. The geometric alignment is determined based on a preset residual threshold. If the alignment requirements are met, the current pose compensation amount is solidified as a multi-camera consistency calibration parameter.

[0174] Additionally, embodiments of the present invention provide a multi-camera acquisition method. Figure One Consistency calibration system, including:

[0175] The coarse matching module is used to acquire reference images of the calibration board fixed on the detection platform by multiple cameras, and to perform coarse matching analysis based on geometric feature distance on the reference images. If the preset matching conditions are not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters.

[0176] The coordinate mapping module is used to drive the target camera in the multi-camera system to perform a predetermined displacement along the orthogonal direction and acquire a displacement image sequence after the coarse matching is passed. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from physical space to image space are obtained.

[0177] The pose compensation module is used to perform arbitrary pose perturbation on the target camera and acquire the perturbation image. By analyzing the affine transformation parameters of the reference image and the perturbation image and combining them with the coordinate mapping parameters, the pose compensation amount, which includes the displacement compensation amount and the rotation correction angle, is obtained.

[0178] The camera consistency calibration module is used to drive the target camera to perform pose compensation operations including reverse displacement and rotation based on the pose compensation amount. It acquires a verification image set through the compensated multi-camera imaging space and generates multi-camera consistency calibration parameters based on the geometric feature alignment of the verification image set.

[0179] Furthermore, embodiments of the present invention provide a multi-camera acquisition method. Figure One A consistency calibration device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the multi-camera acquisition described above. Figure One Consistency calibration method. This equipment provides a highly reliable multi-camera calibration solution for precision manufacturing scenarios such as display modules through deep hardware and software co-design and dynamic error compensation mechanism.

[0180] Meanwhile, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the multi-camera acquisition described above. Figure One A consistency calibration method. The method includes a complete calibration process: dynamic threshold coarse matching, orthogonal displacement mapping modeling, affine parameter compensation analysis, and geometric alignment verification. The storage medium includes, but is not limited to, solid-state drives, optical discs, or embedded flash memory, and its technical features, together with the method claims, form a hardware-software collaborative technical protection system.

[0181] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.

[0182] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0183] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0184] It should be noted that any reference numerals placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims that enumerate several means, several of these means may be embodied by the same hardware. The use of the terms first, second, third, etc., is merely for convenience of expression and does not indicate any order. These terms can be understood as part of the component names.

[0185] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0186] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the claims should be interpreted to include both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0187] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention should also include these modifications and variations.

Claims

1. A method for calibrating the consistency of multi-camera image capture in display module manufacturing, characterized in that, include: Reference images of the calibration board fixed on the detection platform are acquired by multiple cameras. Coarse matching analysis based on geometric feature distance is performed on the reference images. If the preset matching conditions are not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters. After the coarse matching is passed, the target camera in the multi-camera system is driven to perform a predetermined displacement along the orthogonal direction and acquire a displacement image sequence for the display module. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from physical space to image space are obtained. Arbitrary pose perturbation is applied to the target camera and perturbation images are acquired. The pose compensation amount, which includes displacement compensation amount and rotation correction angle, is obtained by analyzing the affine transformation parameters of the reference image and the perturbation image and combining the coordinate mapping parameters. Based on the pose compensation amount, the target camera is driven to perform pose compensation operations including reverse displacement and rotation. A verification image set is acquired through the compensated multi-camera imaging space, and multi-camera consistency calibration parameters are generated based on the geometric feature alignment of the verification image set.

2. The multi-camera image capture consistency calibration method for display module manufacturing as described in claim 1, characterized in that, Reference images of a calibration board fixed to the detection platform are acquired by multiple cameras. Coarse matching analysis based on geometric feature distance is performed on the reference images. If the preset matching conditions are not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters, including: The first and second reference images are acquired by the upper and lower cameras respectively, which are fixed to the calibration plate of the detection platform. Sub-pixel level coordinates of feature points in each baseline image are extracted to generate candidate point pairs from multiple camera perspectives. The Euclidean distance between each pair of feature points is calculated to obtain the symmetric distance matrix between the reference images. Cross-camera similarity comparison is then performed on the distance matrix elements in each row of the symmetric distance matrix. When each row of the distance matrix element in the symmetric distance matrix between reference images satisfies the preset coarse matching condition, it is marked as a coarse matching point pair; When the distance matrix elements in each row of the symmetric distance matrix between reference images do not meet the preset coarse matching conditions, the coarse matching parameters are reconstructed by dynamically adjusting the distance tolerance threshold or by rotating the calibration plate / camera plane through a linkage mechanical mechanism until the preset coarse matching conditions are met. in, The upper camera is detachably mounted at the end of the drive mechanism, and the drive mechanism enables adjustment of at least two-dimensional planar translation and rotation around the optical axis, with the optical axis pointing towards the upper surface of the object being detected. The lower camera is an imaging device that is fixedly installed below the detection platform, with its optical axis perpendicular to the platform's bearing surface and pointing towards the lower surface of the object being detected; The calibration board is equipped with a geometric feature array, the spacing between each feature point is known and the arrangement pattern is locally unique; The coarse matching condition is as follows: cross-matrix row comparison is performed in the symmetric distance matrix between the reference images. When there are at least three elements in the distance element set of a certain row of one symmetric distance matrix and the corresponding distance element set of a certain row of another symmetric distance matrix whose difference is less than a preset threshold, the feature point pair corresponding to the two rows is determined to be a coarse matching point pair.

3. The multi-camera image acquisition consistency calibration method for display module manufacturing as described in claim 2, characterized in that, After successful matching, and after successful coarse matching, the target camera in the multi-camera system is driven to perform a predetermined displacement along orthogonal directions and acquire a displacement image sequence for the display module. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from physical space to image space are obtained, including: The upper camera in the multi-camera system is controlled by a preset drive mechanism to move a first predetermined physical displacement along a first orthogonal direction, and a first orthogonal displacement image is acquired. Based on the matching point set of the second reference image set and the first orthogonal displacement image, a robust estimation algorithm is used to parse the first matching matrix containing the displacement component. The camera is controlled by a preset drive mechanism to move a predetermined displacement along the second orthogonal direction and acquire a second displacement image. Based on the matching point set of the first displacement image and the second displacement image, a robust estimation algorithm is used to parse the second matching matrix containing the displacement component. Based on the mapping relationship between the first predetermined physical displacement and the first matching matrix, a spatial constraint equation for the first orthogonal direction is established. Based on the mapping relationship between the second predetermined physical displacement and the second matching matrix, establish the spatial constraint equation for the second orthogonal direction; By simultaneously solving the spatial constraint equations of the first orthogonal direction and the spatial constraint equations of the second orthogonal direction, the coordinate mapping parameters from the physical space to the image space are calculated.

4. The multi-camera image acquisition consistency calibration method for display module manufacturing as described in claim 3, characterized in that, The first matching matrix is: In the formula, H1 is the first matching matrix, and the components are... This indicates that a pixel displacement of magnitude Δu1 is generated in the horizontal direction of the image coordinate system, and a pixel displacement of magnitude Δv1 is generated in the vertical direction of the image coordinate system. The spatial constraint equation for the first orthogonal direction is: In the formula, Let T represent the displacement of the camera in the first orthogonal direction of the physical coordinate system as 'a', and the displacement in the second orthogonal direction of the physical coordinate system as 0. 00 T is the mapping scaling factor of the physical displacement in the first orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system. 10 T is the scaling factor for the physical displacement in the first orthogonal direction of the physical coordinate system to the perpendicular direction of the image coordinate system. 01 T is the scaling factor for the physical displacement in the second orthogonal direction of the physical coordinate system to the horizontal direction of the image coordinate system. 11 This is the scaling factor for the physical displacement in the second orthogonal direction of the physical coordinate system to the vertical direction of the image coordinate system; The second matching matrix is: In the formula, H2 is the second matching matrix, and the components are... This indicates that a pixel displacement of magnitude Δu2 is generated in the horizontal direction of the image coordinate system, and a pixel displacement of magnitude Δv2 is generated in the vertical direction of the image coordinate system. The spatial constraint equation for the second orthogonal direction is: In the formula, This indicates that the camera's displacement in the first orthogonal direction of the physical coordinate system is 0, and its displacement in the second orthogonal direction of the physical coordinate system is b. The translation mapping parameters are:

5. The multi-camera image acquisition consistency calibration method for display module manufacturing as described in claim 3, characterized in that, Arbitrary pose perturbation is applied to the target camera and perturbation images are acquired. The pose compensation, including displacement compensation and rotation correction angles, is obtained by analyzing the affine transformation parameters of the reference image and the perturbation image, combined with coordinate mapping parameters. The camera is controlled by a preset drive mechanism to generate arbitrary pose disturbances, and the disturbance image after the pose disturbance is acquired. Based on the matching point set between the perturbed image and the first reference image, a robust estimation algorithm is used to parse the third matching matrix containing displacement and rotation components. Based on the translation components and coordinate mapping parameters of the third matching matrix, the displacement compensation amount in the physical space is calculated. The third matching matrix is ​​normalized using a pre-calibrated camera intrinsic parameter matrix, and then the normalized transformation matrix is ​​decomposed into rotation components to obtain the rotation correction angle. By fusing displacement compensation and rotation correction angle, pose compensation for multi-camera pose consistency is generated.

6. The multi-camera image acquisition consistency calibration method for display module manufacturing as described in claim 5, characterized in that, The third matching matrix is: In the formula, H3 is the third matching matrix, h 11 h represents the contribution of the horizontal displacement of the physical coordinate system to the scaling and rotation of the image coordinate system in the horizontal direction. 12 h represents the contribution of the vertical displacement of the physical coordinate system to the shearing and rotation of the horizontal direction of the image coordinate system. 21 h represents the contribution of the horizontal displacement of the physical coordinate system to the shearing and rotation in the vertical direction of the image coordinate system. 22 h represents the contribution of the displacement in the vertical direction of the physical coordinate system to the scaling and rotation in the vertical direction of the image coordinate system. 13 h represents the horizontal displacement component in the image coordinate system. 23 h represents the vertical displacement component in the image coordinate system. 31 h represents the horizontal projection component of the perspective distortion caused by the tilt of the calibration plate / camera. 32 The vertical projection component of the perspective distortion caused by the tilt of the calibration plate / camera; The displacement compensation amount in physical space is: In the formula, x1 is the displacement that the upper camera needs to compensate for in the horizontal direction of the physical coordinate system, and y1 is the displacement that the upper camera needs to compensate for in the vertical direction of the physical coordinate system. Normalized transformation matrix: H norm =K -1 *H3*K; In the formula, H norm Here, K is the normalized transformation matrix, and K is the pre-calibrated camera intrinsic parameter matrix. The rotation correction angle is: In the formula, C is the rotation correction angle, and H norm [0,0] represents the element in the first row and first column of the normalized matrix, H norm [1,0] represents the element in the second row and first column of the normalized matrix.

7. The multi-camera image acquisition consistency calibration method for display module manufacturing as described in claim 5, characterized in that, The first matching matrix, the second matching matrix, and the third matching matrix are all obtained through the following steps: Extract sub-pixel level feature point coordinates from the image pairs to be matched, and generate candidate point pairs from multiple camera perspectives; The Euclidean distance between each pair of feature points is calculated to obtain the symmetric distance matrix between the pairs of images to be matched. Cross-camera similarity comparison is then performed on the distance matrix elements in each row of the symmetric distance matrix. When each row of the distance matrix element in the symmetric distance matrix satisfies the preset coarse matching condition, it is marked as a coarse matching point pair; When the distance matrix elements in each row of the symmetric distance matrix do not meet the preset coarse matching conditions, the coarse matching parameters are reconstructed by dynamically adjusting the distance tolerance threshold or by rotating the calibration plate / camera plane through a linkage mechanical mechanism until the preset coarse matching conditions are met. A polar coordinate neighborhood is determined with each candidate point as the center. Neighboring points are selected in the polar coordinate neighborhood and the polar angles of each neighborhood point relative to the candidate points are calculated. After sorting, a polar angle sequence representing the distribution relationship of neighborhood angles across camera views is obtained. Perform differential angle transformation on the polar angle sequence across camera views to eliminate rotational bias and generate the obtained differential angle sequence; Valid matching point pairs are determined by comparing the similarity of the differential angle sequences. When the similarity of the differential angle sequences exceeds the second threshold, they are determined to be valid matching point pairs. If there are multiple valid matching point pairs, the point pair with the highest similarity in neighborhood geometric structure is selected to form the final matching point set. The homography matrix is ​​solved by a combination algorithm of random sampling consistency and least squares method for the final matching point set to obtain the first matching matrix, the second matching matrix, or the third matching matrix. Specifically, when calculating the first matching matrix, the images to be matched are the second reference image and the first orthogonal displacement image; when calculating the second matching matrix, the images to be matched are the first displacement image and the second displacement image; and when calculating the third matching matrix, the images to be matched are the perturbation image and the first reference image.

8. A multi-camera image acquisition consistency calibration system for display module manufacturing, characterized in that, include: The coarse matching module is used to acquire reference images of the calibration board fixed on the detection platform by multiple cameras, and to perform coarse matching analysis based on geometric feature distance on the reference images. If the preset matching conditions are not met, the matching threshold is dynamically adjusted or the spatial pose of the rotating calibration board / camera is adjusted to reconstruct the coarse matching parameters. The coordinate mapping module is used to drive the target camera in the multi-camera system to perform a predetermined displacement along the orthogonal direction after the coarse matching is passed, and to acquire the displacement image sequence for the display module. Based on the spatial transformation relationship of the displacement image sequence in different orthogonal directions, the coordinate mapping parameters from physical space to image space are obtained. The pose compensation module is used to perform arbitrary pose perturbation on the target camera and acquire the perturbation image. By analyzing the affine transformation parameters of the reference image and the perturbation image and combining the coordinate mapping parameters, the pose compensation amount, which includes the displacement compensation amount and the rotation correction angle, is obtained. The camera consistency calibration module is used to drive the target camera to perform pose compensation operations including reverse displacement and rotation based on the pose compensation amount. It acquires a verification image set through the compensated multi-camera imaging space and generates multi-camera consistency calibration parameters based on the geometric feature alignment of the verification image set.

9. A multi-camera image acquisition consistency calibration device, characterized in that, include: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which are executed by at least one processor to enable the at least one processor to perform the multi-camera image acquisition consistency calibration method for display module manufacturing as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer-executable instructions thereon, characterized in that, When the executable instructions are executed by the processor, they implement the multi-camera image capture consistency calibration method for display module manufacturing as described in any one of claims 1-7.

Images