A multi-camera high-precision plane vision system calibration method and system
By automatically generating planar calibration board graphics and combining coarse and fine alignment calibration methods, the problem of insufficient calibration flexibility and robustness of multi-camera high-precision planar vision systems is solved, achieving accurate, easy-to-use, and reliable extrinsic parameter calibration, and reducing labor and economic costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN WENTU TECHNOLOGY CO LTD
- Filing Date
- 2024-12-16
- Publication Date
- 2026-05-19
AI Technical Summary
Existing calibration methods for multi-camera high-precision planar vision systems are insufficient in terms of flexibility and robustness, leading to calibration failures or errors in different application scenarios, and increasing labor and economic costs.
By automatically generating planar calibration board graphics and combining coarse and fine alignment calibration methods, the extrinsic parameters of each camera and the coordinate transformation from image space to calibration board graphic model space are determined, including feature point generation, convex hull point set feature vector matching, and k-nearest neighbor feature vector selection.
It enables accurate, easy-to-use, and reliable extrinsic parameter calibration of multi-camera high-precision planar vision systems, improving the flexibility, accuracy, and reliability of calibration while reducing labor and economic costs.
Smart Images
Figure CN119832086B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of machine vision inspection, and specifically to a calibration method and system for a multi-camera high-precision planar vision system. Background Technology
[0002] A multi-camera high-precision planar vision system is a system that utilizes multiple cameras for image acquisition and processing, aiming to improve the accuracy and efficiency of planar measurement and inspection. This system typically includes multiple cameras, a calibration board, data processing algorithms, and other components, which work together to achieve high-precision planar vision measurement and inspection. Through multi-camera collaborative operation, efficient calibration methods, and advanced data processing algorithms, multi-camera high-precision planar vision systems can significantly improve the accuracy and efficiency of planar measurement and inspection, and are widely used in various fields such as industrial inspection, autonomous driving, and scientific research.
[0003] In multi-camera high-precision planar vision systems, maintaining consistent shooting postures (including height and angle) across all cameras is crucial for image capture. Furthermore, image data processing requires transforming the target coordinates from each camera's image space to a unified coordinate system. Both of these aspects necessitate calibration. Current calibration methods for multi-camera high-precision planar vision systems suffer from insufficient flexibility and robustness in application.
[0004] 1. There are significant differences in the number of cameras, overall field of view, and single-pixel accuracy of multi-camera high-precision planar vision systems under different application scenarios. Calibration boards designed for a specific multi-camera high-precision planar vision system often fail in other systems, and there is a lack of simple and universal calibration board design methods.
[0005] 2. During the camera installation and debugging phase of a multi-camera high-precision planar vision system, there may be significant differences in the shooting posture of each camera, leading to calibration failures or incorrect calibration results.
[0006] The aforementioned problems have led to high labor costs and economic burdens in the production and use of equipment incorporating multi-camera high-precision planar vision systems. Therefore, it is necessary to propose a new calibration method for multi-camera high-precision planar vision systems to address these two prominent issues. Summary of the Invention
[0007] This invention addresses the shortcomings of current multi-camera high-precision planar vision system calibration methods in terms of flexibility and robustness by providing a calibration method and system for multi-camera high-precision planar vision systems, achieving accurate, easy-to-use, and reliable multi-camera extrinsic parameter calibration.
[0008] The first aspect of this invention relates to a calibration method for a multi-camera high-precision planar vision system, used to determine the extrinsic parameters of each camera in the system and the coordinate transformation from its image space to the calibration board graphic model space, specifically including the following steps:
[0009] S1: Generate a planar calibration board graphic using a calibration board graphic generation method based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system.
[0010] S2: Create a planar calibration board based on the planar calibration board graphic, and use the cameras of the planar vision system to capture images of the planar calibration board to obtain the planar calibration board image;
[0011] S3: For the planar calibration board image and calibration board graphic, use the coarse alignment calibration method to determine the coordinate transformation from the image space of each camera in the planar vision system to the graphic model space of the calibration board;
[0012] S4: For the coordinate transformation from the image space to the graphic model space of each camera in the planar vision system obtained from the planar calibration plate image, calibration plate graphic, and coarse alignment calibration, the fine alignment calibration method is used to determine the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the graphic model space of the calibration plate.
[0013] According to the above technical solution, the specific steps of the calibration board pattern generation method in S1 are as follows:
[0014] S1.1: Based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system, determine the calibration board size, feature point radius, feature point array size, and minimum distance d between nearest neighbor feature points. min The maximum distance d between nearest neighbor feature points max The calibration plate graphic generation parameters include the number of feature points contained in the feature point matrix;
[0015] S1.2: Randomly generate a set of feature points within the feature point matrix based on the number of feature points contained in the feature point matrix;
[0016] S1.3: Traverse each feature point in the feature point set, calculate its distance to its nearest neighbor feature point, and remove feature points whose distance to their nearest neighbor is less than the minimum distance d. min Feature points;
[0017] S1.4: Check if the distance between the nearest neighbor feature points of all feature points in the feature point set is greater than the maximum distance d between the nearest neighbor feature points. max If yes, proceed to S1.2; otherwise, continue to S1.5.
[0018] S1.5: Calculate the convex hull of the feature point set to obtain the convex hull point set containing n feature points;
[0019] S1.6: Traverse each feature point in the convex hull point set, and record the n side lengths of the convex hull polygon starting from that point in a counterclockwise direction to obtain a feature vector set containing n n-dimensional feature vectors;
[0020] S1.7: Check the distance between any two n-dimensional feature vectors in the feature vector set. If there are two n-dimensional feature vectors whose similarity is greater than a preset threshold... If the condition is met, proceed to S1.2; otherwise, continue to S1.8.
[0021] S1.8: Copy and translate the feature point array composed of feature point sets, with multiple feature point arrays moving at a distance greater than d. max The fixed-interval horizontal and vertical arrangement of feature dots forms a grid that covers the entire planar calibration plate;
[0022] S1.9: Add a unique ID identifier to each feature point in the feature point grid to form a planar calibration plate graphic.
[0023] According to the above technical solution, the specific steps of the coarse alignment calibration method in S3 are as follows:
[0024] S3.1: Select a calibration board image taken by a camera, and extract the center of all feature points in the image to obtain the feature center point set;
[0025] S3.2: Taking the points in the feature center set as vertices, where the distance is less than d. max Adding edges to point pairs forms a feature map;
[0026] S3.3: Extract connected components from the feature map. Each connected component corresponds to a feature point matrix. Identify the ID of the feature point matrix from the image. Discard connected components whose feature point matrix IDs fail to be identified. Discard connected components whose number of vertices is not equal to the number of feature points in the feature point matrix.
[0027] S3.4: Select a feature point matrix that has been successfully identified and has not been discarded, extract the convex hull point set of the feature center point set of the feature point matrix, traverse each feature point in the convex hull point set, and record the n side lengths of the convex hull polygon starting from the point in a counterclockwise direction to obtain the calibration board image feature vector set containing n n-dimensional feature vectors.
[0028] S3.5: Extract the convex hull point set of the feature point set matching the feature point matrix ID in the calibration board graphic, traverse each feature point in the convex hull point set, and record the n side lengths of the convex hull polygon starting from that point in a counterclockwise direction to obtain the calibration board graphic feature vector set containing n n-dimensional feature vectors.
[0029] S3.6: Perform pairwise matching of the feature vector set of the feature calibration board image with the feature vector set of the calibration board graphic to calculate the similarity, and determine the maximum feature vector similarity as the current feature point matrix image-graphic similarity;
[0030] S3.7: Repeat S3.4 to 3.6 until all successfully identified and not discarded feature points are traversed, and determine the feature point ID with the highest feature point image-graphic similarity;
[0031] S3.8: Based on the feature point ID of the maximum feature point image-graphic similarity, determine the matching relationship between its feature center point set and feature point set, and calculate the coordinate transformation from the current camera's image space to the calibration board's graphic model space based on the matched point pairs.
[0032] S3.9: Repeat S3.1 to S3.8 until all the calibration board images captured by the cameras are traversed, and obtain the coordinate transformation from the image space of each camera of the planar vision system to the graphic model space of the calibration board.
[0033] According to the above technical solution, the specific steps of the fine alignment calibration method in S4 are as follows:
[0034] S4.1: For each feature point in the calibration board graphic feature point set, find its k nearest neighbor feature points in the calibration board graphic feature point set, calculate the unit direction vector of the plane vector formed by the feature point and each of its corresponding k nearest neighbor mapping feature center points, sort the unit direction vectors by direction and concatenate them into a 2k-dimensional k nearest neighbor feature vector.
[0035] S4.2: Select a calibration board image captured by a camera. Based on the coordinate transformation from the current camera's image space to the calibration board's graphic model space obtained from coarse alignment calibration, map each feature center point in the extracted feature center point set to the calibration board's graphic model space to obtain the mapped feature center point set.
[0036] S4.3: For each mapping feature center point in the set of mapping feature center points, find its k nearest neighbor mapping feature center points in the set of mapping feature center points, calculate the unit direction vector of the plane vector formed by the mapping feature center point and each of its corresponding k nearest neighbor mapping feature center points, sort the unit direction vectors according to their directions and concatenate them into a 2k-dimensional k nearest neighbor feature vector.
[0037] S4.4: For each mapping feature center point in the set of mapping feature center points, find the calibration board graphic feature point set whose distance to it is less than a preset threshold d. c The feature points of the calibration board graphic are used to form a set of candidate matching feature points for the center point of the mapping feature;
[0038] S4.5: For each mapped feature center point in the mapped feature center point set, calculate the similarity between its k-nearest neighbor feature vector and the k-nearest neighbor feature vectors of each candidate matching feature point in the corresponding candidate matching feature point set. Determine the candidate matching feature point with the highest similarity as the best matching feature point, and discard those with similarity below a preset threshold. The matching results of the center points of the mapping features are obtained, and the matching results of the set of center points of the mapping features and the set of feature points of the calibration board are obtained.
[0039] S4.6: Based on the matching results of the mapping feature center point set and the calibration board graphic feature point set, calculate the matching results of the calibration board image feature center point set and the calibration board graphic feature point set;
[0040] S4.7: Based on the matching results between the feature center point set of the calibration board image and the feature point set of the calibration board graphic, calculate the coordinate transformation from the current camera's image space to the calibration board graphic model space;
[0041] S4.8: Repeat S4.2 to S4.7 until all the calibration board images captured by the cameras are traversed, to obtain the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the calibration board graphic model space.
[0042] A second aspect of the present invention relates to a multi-camera high-precision planar vision system, comprising the above-described multi-camera high-precision planar vision system calibration method, a memory, and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code.
[0043] This invention automatically generates a planar calibration board graphic based on the design parameters of a high-precision planar vision system using a planar calibration board graphic generation method. By sequentially performing coarse and fine alignment calibration processes, it achieves accurate, easy-to-use, and reliable multi-camera extrinsic parameter calibration. The multi-camera high-precision planar vision system calibration method disclosed in this invention has three beneficial effects:
[0044] This invention uses a calibration plate graphic generation method to automatically generate planar calibration plate graphics based on the design parameters of a multi-camera high-precision planar vision system. Applicable planar calibration plate graphics can be automatically generated for multi-camera high-precision planar vision systems in different application scenarios, improving the flexibility and ease of use of the calibration method.
[0045] This invention uses a coarse alignment calibration method to perform coarse-grained alignment matching and calibration, and constructs feature vectors of the convex hull point set for feature matching. This can avoid calibration failure or errors caused by pose differences of multiple cameras, and improve the accuracy and reliability of calibration.
[0046] This invention performs precise alignment and calibration using a fine alignment calibration method. It constructs k-nearest neighbor feature vectors for feature matching, compares and filters candidate matches, and utilizes all successfully matched feature points within the camera's field of view for calibration calculation, further improving the accuracy and precision of the calibration.
[0047] Based on the above three aspects, this invention achieves accurate, easy-to-use, and reliable multi-camera extrinsic parameter calibration for multi-camera high-precision planar vision systems. Attached Figure Description
[0048] Figure 1 This is a schematic diagram of the planar calibration plate of the present invention.
[0049] Figure 2 This is a schematic diagram illustrating the convex hull point set feature vector composition of a feature point in this invention.
[0050] Figure 3 This is a schematic diagram illustrating the k-nearest neighbor feature vectors of a feature point in this invention.
[0051] Figure 4 This is a schematic diagram of the process of the present invention.
[0052] Figure 5 This is a flowchart illustrating the method for generating calibration board graphics.
[0053] Figure 6 This is a flowchart illustrating the coarse alignment calibration method.
[0054] Figure 7 This is a flowchart illustrating the precise alignment calibration method. Detailed Implementation
[0055] This invention discloses a calibration method for a multi-camera high-precision planar vision system. The specific implementation of this invention will be further described below with reference to preferred embodiments and the accompanying drawings.
[0056] Example 1
[0057] Reference Figures 1-7 The chip bracket appearance defect detection template matching and positioning method of the present invention includes the following steps:
[0058] S1: Generate a planar calibration board graphic using a calibration board graphic generation method based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system.
[0059] S2: Create a planar calibration board based on the planar calibration board graphic, and use the cameras of the planar vision system to capture images of the planar calibration board to obtain the planar calibration board image;
[0060] S3: For the planar calibration board image and calibration board graphic, use the coarse alignment calibration method to determine the coordinate transformation from the image space of each camera in the planar vision system to the graphic model space of the calibration board;
[0061] S4: For the coordinate transformation from the image space to the graphic model space of each camera in the planar vision system obtained from the planar calibration plate image, calibration plate graphic, and coarse alignment calibration, the fine alignment calibration method is used to determine the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the graphic model space of the calibration plate.
[0062] like Figure 1 As shown, the specific steps of the calibration plate pattern generation method in S1 above are as follows:
[0063] S1.1: Based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system, determine the calibration board size, feature point radius, feature point array size, and minimum distance d between nearest neighbor feature points. min The maximum distance d between nearest neighbor feature points max The calibration plate graphic generation parameters include the number of feature points contained in the feature point matrix;
[0064] S1.2: Randomly generate a set of feature points within the feature point matrix based on the number of feature points contained in the feature point matrix;
[0065] S1.3: Traverse each feature point in the feature point set, calculate its distance to its nearest neighbor feature point, and remove feature points whose distance to their nearest neighbor is less than the minimum distance d. min Feature points;
[0066] S1.4: Check if the distance between the nearest neighbor feature points of all feature points in the feature point set is greater than the maximum distance d between the nearest neighbor feature points. max If yes, proceed to S1.2; otherwise, continue to S1.5.
[0067] S1.5: Calculate the convex hull of the feature point set to obtain the convex hull point set containing n feature points;
[0068] S1.6: As Figure 2 As shown, by traversing each feature point in the convex hull point set and recording the n side lengths of the convex hull polygon starting from that point in a counterclockwise direction, a feature vector set containing n n-dimensional feature vectors is obtained.
[0069] S1.7: Check the distance between any two n-dimensional feature vectors in the feature vector set. If there are two n-dimensional feature vectors whose similarity is greater than a preset threshold... If the condition is met, proceed to S1.2; otherwise, continue to S1.8.
[0070] S1.8: Copy and translate the feature point array composed of feature point sets, with multiple feature point arrays moving at a distance greater than d. max The fixed-interval horizontal and vertical arrangement of feature dots forms a grid that covers the entire planar calibration plate;
[0071] S1.9: Add a unique ID identifier to each feature point in the feature point grid to form a planar calibration plate graphic.
[0072] Preferably, the specific method for determining the calibration board graphic generation parameters in S1.1 based on the field of view of the planar vision system, the number of cameras, and the single-pixel accuracy is as follows: the calibration board size is set to be slightly larger than the field of view of the planar vision system; the feature point radius is 1 / 100 of the width or height of the target plane field of view of a single camera; the feature point array size is set to 1 / 3 of the width and height of the target plane field of view of a single camera to ensure that there is at least one complete feature point array within the camera's field of view; and the number of feature points contained in the feature point array is set to the maximum distance d between the nearest neighbor feature points. max The number of densest embedded feature points within the feature point matrix range is given by the minimum distance d between the nearest neighbor feature points. min The value between the number of feature points most densely embedded within the feature point matrix and the number of feature points embedded within the matrix.
[0073] Preferably, the feature points generated in S1.2 are solid circles.
[0074] Preferably, the specific method for calculating the convex hull of the feature point set in S1.5 to obtain the convex hull point set containing n feature points is the Graham scan algorithm.
[0075] Preferably, the specific method for adding a unique ID identifier to each feature dot matrix in S1.9 is to add a numeric character identifier feature array ID within the range of each feature dot matrix.
[0076] The specific steps of the coarse alignment calibration method in S3 above are as follows:
[0077] S3.1: Select a calibration board image taken by a camera, and extract the center of all feature points in the image to obtain the feature center point set;
[0078] S3.2: Taking the points in the feature center set as vertices, where the distance is less than d. max Adding edges to point pairs forms a feature map;
[0079] S3.3: Extract connected components from the feature map. Each connected component corresponds to a feature point matrix. Identify the ID of the feature point matrix from the image. Discard connected components whose feature point matrix IDs fail to be identified. Discard connected components whose number of vertices is not equal to the number of feature points in the feature point matrix.
[0080] S3.4: Select a feature point matrix that has been successfully identified and has not been discarded, extract the convex hull point set of the feature center point set of the feature point matrix, traverse each feature point in the convex hull point set, and record the n side lengths of the convex hull polygon starting from the point in a counterclockwise direction to obtain the calibration board image feature vector set containing n n-dimensional feature vectors.
[0081] S3.5: Extract the convex hull point set of the feature point set matching the feature point matrix ID in the calibration board graphic, traverse each feature point in the convex hull point set, and record the n side lengths of the convex hull polygon starting from that point in a counterclockwise direction to obtain the calibration board graphic feature vector set containing n n-dimensional feature vectors.
[0082] S3.6: Perform pairwise matching of the feature vector set of the feature calibration board image with the feature vector set of the calibration board graphic to calculate the similarity, and determine the maximum feature vector similarity as the current feature point matrix image-graphic similarity;
[0083] S3.7: Repeat S3.4 to 3.6 until all successfully identified and not discarded feature points are traversed, and determine the feature point ID with the highest feature point image-graphic similarity;
[0084] S3.8: Based on the feature point ID of the maximum feature point image-graphic similarity, determine the matching relationship between its feature center point set and feature point set, and calculate the coordinate transformation from the current camera's image space to the calibration board's graphic model space based on the matched point pairs.
[0085] S3.9: Repeat S3.1 to S3.8 until all the calibration board images captured by the cameras are traversed, and obtain the coordinate transformation from the image space of each camera of the planar vision system to the graphic model space of the calibration board.
[0086] Preferably, the specific method for extracting the center of all feature points in the image in S3.1 is as follows: a binary image is obtained by thresholding using the Otsu method; connected components are extracted from the binary image to obtain the region where each feature point is located; and the centroid method is used to extract the center of feature points at the sub-pixel level.
[0087] The specific steps of the precise alignment calibration method in S4 above are as follows:
[0088] S4.1: As Figure 3 As shown, for each feature point in the calibration board graphic feature point set, find its k nearest neighbor feature points in the calibration board graphic feature point set, calculate the unit direction vector of the plane vector formed by the feature point and each of its corresponding k nearest neighbor mapping feature center points, sort the unit direction vectors according to their directions and concatenate them into a 2k-dimensional k nearest neighbor feature vector.
[0089] S4.2: Select a calibration board image captured by a camera. Based on the coordinate transformation from the current camera's image space to the calibration board's graphic model space obtained from coarse alignment calibration, map each feature center point in the extracted feature center point set to the calibration board's graphic model space to obtain the mapped feature center point set.
[0090] S4.3: For each mapping feature center point in the set of mapping feature center points, find its k nearest neighbor mapping feature center points in the set of mapping feature center points, calculate the unit direction vector of the plane vector formed by the mapping feature center point and each of its corresponding k nearest neighbor mapping feature center points, sort the unit direction vectors according to their directions and concatenate them into a 2k-dimensional k nearest neighbor feature vector.
[0091] S4.4: For each mapping feature center point in the set of mapping feature center points, find the calibration board graphic feature point set whose distance to it is less than a preset threshold d. c The feature points of the calibration board graphic are used to form a set of candidate matching feature points for the center point of the mapping feature;
[0092] S4.5: For each mapped feature center point in the mapped feature center point set, calculate the similarity between its k-nearest neighbor feature vector and the k-nearest neighbor feature vectors of each candidate matching feature point in the corresponding candidate matching feature point set. Determine the candidate matching feature point with the highest similarity as the best matching feature point, and discard those with similarity below a preset threshold. The matching results of the center points of the mapping features are obtained, and the matching results of the set of center points of the mapping features and the set of feature points of the calibration board are obtained.
[0093] S4.6: Based on the matching results of the mapping feature center point set and the calibration board graphic feature point set, calculate the matching results of the calibration board image feature center point set and the calibration board graphic feature point set;
[0094] S4.7: Based on the matching results between the feature center point set of the calibration board image and the feature point set of the calibration board graphic, calculate the coordinate transformation from the current camera's image space to the calibration board graphic model space;
[0095] S4.8: Repeat S4.2 to S4.7 until all the calibration board images captured by the cameras are traversed, to obtain the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the calibration board graphic model space.
[0096] Example 2:
[0097] This embodiment relates to a multi-camera high-precision planar vision system calibration method and apparatus, including a memory and one or more processors. The memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the multi-camera high-precision planar vision system calibration method of this embodiment 1.
[0098] Example 3:
[0099] This embodiment relates to a computer-readable storage medium storing a program that, when executed by a processor, implements a multi-camera high-precision planar vision system calibration method of Embodiment 1.
[0100] The embodiments described in this specification are merely examples of implementations of the inventive concept. The scope of protection of this invention should not be considered as limited to the specific forms described in the embodiments. The scope of protection of this invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.
Claims
1. A calibration method for a multi-camera high-precision planar vision system, characterized in that: S1: Generate a planar calibration board graphic based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system using a calibration board graphic generation method; S2: Create a planar calibration board based on the planar calibration board graphic, and use the various cameras of the planar vision system to capture images of the planar calibration board to obtain the planar calibration board image; S3: For the planar calibration board image and calibration board graphic, use the coarse alignment calibration method to determine the coordinate transformation from the image space of each camera in the planar vision system to the graphic model space of the calibration board; S4: For the coordinate transformation from the image space to the graphic model space of each camera in the planar vision system obtained by coarse alignment calibration, the planar calibration plate image, the calibration plate graphic, and the coarse alignment calibration, the fine alignment calibration method is used to determine the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the graphic model space of the calibration plate. The method for generating the calibration board pattern of S1 specifically includes: S1.1: Based on the field of view, number of cameras, and single-pixel accuracy of the planar vision system, determine the calibration board size, feature point radius, feature point array size, and minimum distance d between nearest neighbor feature points. min The maximum distance d between nearest neighbor feature points max The calibration plate graphic generation parameters include the number of feature points contained in the feature point matrix; S1.2: Randomly generate a set of feature points within the feature point matrix based on the number of feature points contained in the feature point matrix; S1.3: Traverse each feature point in the feature point set, calculate its distance to its nearest neighbor feature point, and remove feature points whose distance to their nearest neighbor is less than the minimum distance d. min Feature points; S1.4: Check if the distance between the nearest neighbor feature points of all feature points in the feature point set is greater than the maximum distance d between the nearest neighbor feature points. max If yes, proceed to S1.2; otherwise, continue to S1.
5. S1.5: Calculate the convex hull of the feature point set to obtain the inclusion factor. The set of convex hull points of feature points; S1.6: Traverse each feature point in the convex hull set, and record the convex hull polygon starting from that point in a counter-clockwise direction. The length of the segment is included indivual The set of eigenvectors of a dimensional eigenvector; S1.7: Examine the feature vector set The distance between any two dimensional feature vectors, if there exist two The similarity of the 3D feature vectors is greater than a preset threshold. n If the condition is met, proceed to S1.2; otherwise, continue to S1.
8. S1.8: Copy and translate the feature point array composed of feature point sets, with multiple feature point arrays moving at a distance greater than d. max The fixed-interval horizontal and vertical arrangement of feature dots forms a grid that covers the entire planar calibration plate; S1.9: Add a unique ID identifier to each feature point in the feature point grid to form a planar calibration plate graphic.
2. The calibration method for a multi-camera high-precision planar vision system according to claim 1, characterized in that: The coarse alignment calibration method of S3 specifically includes: S3.1: Select a calibration board image taken by a camera, and extract the center of all feature points in the image to obtain the feature center point set; S3.2: Taking the points in the feature center set as vertices, where the distance is less than d. max Adding edges to point pairs forms a feature map; S3.3: Extract connected components from the feature map. Each connected component corresponds to a feature point matrix. Identify the ID of the feature point matrix from the image. Discard connected components whose feature point matrix IDs fail to be identified. Discard connected components whose number of vertices is not equal to the number of feature points in the feature point matrix. S3.4: Select a feature point matrix that was successfully identified and not discarded, extract the convex hull point set of the feature center point set of the feature point matrix, traverse each feature point in the convex hull point set, and record the convex hull polygon starting from that point in a counterclockwise direction. The length of the segment is included indivual A calibration plate image feature vector set with 3D feature vectors; S3.5: Extract the convex hull point set of the feature point set matching the feature point matrix ID in the calibration board image, traverse each feature point in the convex hull point set, and record the convex hull polygon starting from that point in a counterclockwise direction. The length of the segment is included indivual A set of eigenvectors for a calibration board of eigenvectors; S3.6: Perform pairwise matching of the feature vector set of the feature calibration board image with the feature vector set of the calibration board graphic to calculate the similarity, and determine the maximum feature vector similarity as the current feature point matrix image-graphic similarity; S3.7: Repeat S3.4 to 3.6 until all successfully identified and not discarded feature points are traversed, and determine the feature point ID with the highest feature point image-graphic similarity; S3.8: Based on the feature point ID of the maximum feature point image-graphic similarity, determine the matching relationship between its feature center point set and feature point set, and calculate the coordinate transformation from the current camera's image space to the calibration board's graphic model space based on the matched point pairs. S3.9: Repeat S3.1 to S3.8 until all the calibration board images captured by the cameras are traversed, and obtain the coordinate transformation from the image space of each camera of the planar vision system to the graphic model space of the calibration board.
3. The calibration method for a multi-camera high-precision planar vision system according to claim 1, characterized in that: The precise alignment calibration method of S4 specifically includes: S4.1: For each feature point in the calibration board graphic feature point set, find its position within the calibration board graphic feature point set. For each nearest neighbor feature point, calculate the sum of the sums ... The unit direction vector of the planar vector formed by the nearest neighbor feature center points is obtained by sorting the unit direction vectors by direction and concatenating them. dimension Nearest neighbor feature vectors; S4.2: Select a calibration board image captured by a camera. Based on the coordinate transformation from the current camera's image space to the calibration board's graphic model space obtained from coarse alignment calibration, map each feature center point in the extracted feature center point set to the calibration board's graphic model space to obtain the mapped feature center point set. S4.3: For each mapped feature center point in the set of mapped feature center points, find its position in the set of mapped feature center points. Nearest neighbor mapping feature center points, calculate the mapping feature center point and its corresponding each The unit direction vector of the planar vector formed by the nearest neighbor feature center points is obtained by sorting the unit direction vectors by direction and concatenating them. dimension Nearest neighbor feature vectors; S4.4: For each mapping feature center point in the mapping feature center point set, find the calibration board graphic feature point set whose distance to it is less than a preset threshold. The feature points of the calibration board graphic are used to form a set of candidate matching feature points for the center point of the mapping feature; S4.5: For each mapped feature center point in the set of mapped feature center points, calculate its... Nearest neighbor feature vectors and each candidate matching feature point in the corresponding candidate matching feature point set The similarity of nearest neighbor feature vectors is used to determine the candidate matching feature point with the highest similarity, and discards those with similarity below a preset threshold. k The matching results of the center points of the mapping features are obtained, and the matching results of the set of center points of the mapping features and the set of feature points of the calibration board are obtained. S4.6: Based on the matching results of the mapping feature center point set and the calibration board graphic feature point set, calculate the matching results of the calibration board image feature center point set and the calibration board graphic feature point set; S4.7: Based on the matching results between the feature center point set of the calibration board image and the feature point set of the calibration board graphic, calculate the coordinate transformation from the current camera's image space to the calibration board graphic model space; S4.8: Repeat S4.2 to S4.7 until all the calibration board images captured by the cameras are traversed, to obtain the extrinsic parameters of each camera in the planar vision system and the coordinate transformation from its image space to the calibration board graphic model space.
4. A multi-camera high-precision planar vision system, characterized in that: The system executes the calibration method for a multi-camera high-precision planar vision system as described in claims 1-3. The system includes a memory and one or more processors. The memory stores executable code, and the one or more processors execute the executable code.