A large-size rotating ring precision pose measurement method based on machine vision

Abstract

A kind of large size rotating ring precision pose measurement method based on machine vision, it is characterized by it includes the following steps:1) multi-vision system acquires the image of large size rotating ring;2) accurately extract the edge arc segment key features from the image of rotating ring;3) according to the multi-camera calibration relationship, unify multiple arc segment features to the same coordinate system;4) according to multiple arc segment features and the size of actual rotating ring, the pose of rotating ring is calculated.The present application can realize the function of large size rotating ring precision pose measurement based on machine vision, and solves the problem of large size rotating ring pose precision measurement, and lays a solid foundation for realizing the automatic assembly of large size rotating ring parts.

Description

Technical Field

[0001] This invention relates to a measurement technology, particularly a large-size rotating ring pose measurement technology, specifically a machine vision-based method for accurate pose measurement of a large-size rotating ring. Background Technology

[0002] In the manufacturing process of aerospace products, there are frequent scenarios involving the assembly and docking of large-sized rotating rings. These assembly processes often involve small clearances and short fitting times, thus requiring precise measurement of the rotating ring's position and orientation. Currently, rotating ring docking in China primarily relies on manual operations by operators, depending mainly on their assembly experience. This method is inefficient and prone to collisions and scratches. Therefore, there is an urgent need to develop a precise method for measuring the rotating ring's position and orientation.

[0003] With the deepening research on precision measurement in recent years, visual measurement technology has been widely used due to its non-contact nature, high measurement density, and good system flexibility. However, due to the large size of rotating rings, the field of view of a single camera is insufficient to cover their entire edge features. To address this issue, a machine vision-based method for precise pose measurement of large-size rotating rings is proposed. Summary of the Invention

[0004] This invention addresses the problem of low accuracy in the pose measurement of large-sized rotating rings by proposing a machine vision-based method for precise pose measurement of large-sized rotating rings. This method achieves accurate pose measurement of large-sized rotating rings using machine vision, overcoming the challenge of precise pose measurement for such parts and laying a solid foundation for the automated assembly of large-sized rotating ring components.

[0005] The technical solution of this invention is:

[0006] A machine vision-based method for accurate pose measurement of a large-size rotating ring, characterized by comprising the following steps:

[0007] S1: Use a multi-view vision system to acquire images of a large rotating ring. This multi-view vision system uses four cameras evenly distributed around the rotating ring.

[0008] S2: Accurately extract key features of the edge arc segment from the image of the rotating ring;

[0009] S3: Unify the features of multiple circular arc segments to the same coordinate system based on the multi-view camera calibration relationship;

[0010] S4: Calculate the pose of the rotating ring based on the features of multiple arc segments and the actual size of the rotating ring;

[0011] Step S2, which involves accurately extracting key features of the edge arc segment, includes the following steps:

[0012] S2.1: Image preprocessing of the rotating ring image, including filtering and denoising, image segmentation, and region of interest extraction;

[0013] S2.2: Images acquired by the camera cannot be directly used for calculations because the original images contain a lot of irrelevant information. Therefore, accurate detection of the projected line segments of the rotating ring arc segment spline is required. Key features of the edge arc segment are extracted using the Canny adaptive edge detection and RANSAC-optimized foreground circular imaging ellipse detection method as benchmarks. First, edge detection is performed. The traditional Canny algorithm requires manually setting a fixed double threshold. In contrast, the detected edges are inaccurate. This method uses an adaptive threshold to solve this problem. The acquired image is segmented into foreground and background, where N is the total number of pixels in the image, T is the threshold, α = N1 / N is the proportion of foreground pixels to total pixels, g1 is the average gray level, β = N2 / N is the proportion of background pixels to total pixels, and g2 is the average gray level. The total gray level of the image is denoted as g = α × g1 + β × g2, and the inter-class variance is denoted as α. 2 =α×(g-g1) 2 +β×(g-g2) 2 The maximum threshold T is obtained by optimizing g one by one, and is set as the high threshold in Canny, while T / 2 is set as the low threshold in Canny. In addition, the arc segment is discretized into multiple point sets, and the RANSAC clustering method is used to cluster the point sets, which improves the recognition accuracy and robustness.

[0014] Step S3, which unifies multiple circular arc segment features to the same coordinate system based on the multi-view camera calibration relationship, includes the following steps:

[0015] S3.1: Unifying Camera Coordinates Using a Large Calibration Board. This method uses a large calibration board to unify the coordinates of all cameras. Each large calibration board contains several smaller calibration boards, and the positional relationships between these smaller boards are known. Each camera can capture images of one of these smaller calibration boards. The internal and external parameters of each camera can be calibrated using these smaller calibration boards, and the coordinates of each camera can be transformed to the coordinate system of each smaller calibration board, thus unifying the coordinates of all cameras. The specific calibration process is as follows: Define: The transformation matrices between the camera coordinate systems {C1} and {C2} and the small calibration plate coordinate systems {T1} and {T2} are respectively, and can be obtained by taking pictures of the calibration plate with the camera. This represents the transformation matrix between camera coordinate systems {C1} and {C2}. The transformation matrix between the coordinate systems {T1} and {T2} of the small calibration plate can be obtained from the design of the calibration fixture. Once the coordinate system transformation relationship is determined, the calibration equations can be established. The transformation matrix between camera coordinate systems {C1} and {C2} can be obtained. Similarly, the transformation matrix between the four cameras can be obtained and unified under {C1}.

[0016] S3.2: Unify multiple circular arc features into the same coordinate system based on the calibration parameters of each camera.

[0017] Step S4, which calculates the pose of the rotating ring based on the features of multiple arc segments and the actual dimensions of the rotating ring, includes the following steps:

[0018] S4.1: Fit a complete ellipse feature based on multiple arc segment features;

[0019] S4.2: Calculate the pose of the revolving ring based on the complete elliptical features and the actual dimensions of the revolving ring. After perspective transformation, the spatial circle is imaged as an ellipse on the image plane. This ellipse can be understood as being intercepted by the elliptical cone formed by the image plane, the spatial circle, and the optical center, as shown below. Figure 4 As shown, O w -X w Y w Z w For the world coordinate system, O c -X c Y c Z c O1-X1Y1 is the camera coordinate system, and O1-X1Y1 is the image plane coordinate system. After solving the spatial circle pose in the camera coordinate system, the pose of the circle in the world coordinate system can be calculated by obtaining the extrinsic parameter matrix through camera calibration. Under the premise that the camera has been calibrated, the center position and the pose parameters of the circle plane are determined according to the projection elliptic curve of the circle features on the image. The calculation results are ambiguous. Angle constraints are used to eliminate the ambiguity (see Wei Zhenzhong, Zhao Zheng, Zhang Guangjun: Angle Constraint Elimination of Ambiguity in Spatial Circle Pose Recognition [J]. Optics and Precision Engineering, 2010, 18(03): 685-691).

[0020] The beneficial effects of this invention are:

[0021] This invention enables precise pose measurement of large-sized rotating rings using machine vision, overcoming the challenge of accurate pose measurement of large-sized rotating rings and laying a solid foundation for the automated assembly of large-sized rotating ring parts. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the vision system structure of the present invention.

[0023] Figure 2 This is a schematic diagram of the key features of the extracted edge arc segment involved in step S2 of the embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the multi-view camera calibration involved in step S3 of an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram of the rotation ring pose calculation model involved in step S4 of the embodiment of the present invention.

[0026] Figure 5 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0028] A method for accurate pose measurement of a large-size rotating ring based on machine vision includes the following steps (e.g.) Figure 5 ):

[0029] S1: The multi-view vision system acquires images of a large rotating ring;

[0030] S2: Accurately extract key features of the edge arc segment from the image of the rotating ring;

[0031] S3: Unify the features of multiple circular arc segments to the same coordinate system based on the multi-view camera calibration relationship;

[0032] S4: Calculate the pose of the rotating ring based on the features of multiple arc segments and the actual size of the rotating ring;

[0033] In step S1, the multi-view vision system acquires images of a large rotating ring. A schematic diagram of the overall vision system structure is shown below. Figure 1 As shown, the number of cameras in a multi-view vision system is determined by the size of the rotating ring and the field of view of each camera. The overall field of view of the multi-view vision system should cover the circular arc features of the rotating ring as much as possible. This system uses a four-camera multi-view vision system as an example, with the four cameras evenly distributed around the rotating ring. Figure 1 As shown. Rotating ring 1 follows the end of the assembly platform, while rotating ring 2, as the assembly target, is fixed on the base surface;

[0034] In step S2, key features of the edge arc segment are accurately extracted, such as... Figure 2 As shown, it includes the following steps:

[0035] S2.1: Image preprocessing of the rotating ring image, including filtering and denoising, image segmentation, and region of interest extraction;

[0036] S2.2: Images acquired by the camera cannot be directly used for calculations because the original images contain a lot of irrelevant information. Therefore, accurate detection of the projected line segments of the rotating ring arc segment spline is required. Key features of the edge arc segment are extracted using the Canny adaptive edge detection and RANSAC-optimized foreground circular imaging ellipse detection method as benchmarks. First, edge detection is performed. The traditional Canny algorithm requires manually setting a fixed double threshold. In contrast, the detected edges are inaccurate. This method uses an adaptive threshold to solve this problem. The acquired image is segmented into foreground and background, where N is the total number of pixels in the image, T is the threshold, α = N1 / N is the proportion of foreground pixels to total pixels, g1 is the average gray level, β = N2 / N is the proportion of background pixels to total pixels, and g2 is the average gray level. The total gray level of the image is denoted as g = α × g1 + β × g2, and the inter-class variance is denoted as α. 2 =α×(g-g1) 2 +β×(g-g2) 2 The maximum threshold T is obtained by optimizing g one by one, and is set as the high threshold in Canny, while T / 2 is set as the low threshold in Canny. In addition, the arc segment is discretized into multiple point sets, and the RANSAC clustering method is used to cluster the point sets, which improves the recognition accuracy and robustness.

[0037] In step S3, the features of multiple circular arc segments are unified to the same coordinate system according to the multi-view camera calibration relationship, such as... Figure 3 As shown, it includes the following steps:

[0038] S3.1: Unifying Camera Coordinates Using a Large Calibration Board. This method uses a large calibration board to unify the coordinates of all cameras. Each large calibration board contains several smaller calibration boards, and the positional relationships between these smaller boards are known. Each camera can capture images of one of these smaller calibration boards. The internal and external parameters of each camera can be calibrated using these smaller calibration boards, and the coordinates of each camera can be transformed to the coordinate system of each smaller calibration board, thus unifying the coordinates of all cameras. The specific calibration process is as follows: Define: The transformation matrices between the camera coordinate systems {C1} and {C2} and the small calibration plate coordinate systems {T1} and {T2} are respectively, and can be obtained by taking pictures of the calibration plate with the camera. This represents the transformation matrix between camera coordinate systems {C1} and {C2}. The transformation matrix between the coordinate systems {T1} and {T2} of the small calibration plate can be obtained from the design of the calibration fixture. Once the coordinate system transformation relationship is determined, the calibration equations can be established. The transformation matrix between camera coordinate systems {C1} and {C2} can be obtained. Similarly, the transformation matrix between the four cameras can be obtained and unified under {C1}.

[0039] S3.2: Unify multiple circular arc features into the same coordinate system based on the calibration parameters of each camera.

[0040] In step S4, the pose of the rotating ring is calculated based on the features of multiple arc segments and the actual dimensions of the rotating ring, such as... Figure 4 As shown, it includes the following steps:

[0041] S4.1: Fit a complete ellipse feature based on multiple arc segment features;

[0042] S4.2: Calculate the pose of the revolving ring based on the complete elliptical features and the actual dimensions of the revolving ring. After perspective transformation, the spatial circle is imaged as an ellipse on the image plane. This ellipse can be understood as being intercepted by the elliptical cone formed by the image plane, the spatial circle, and the optical center, as shown below. Figure 4 As shown, O w -X w Y w Z w For the world coordinate system, O c -X c Y c Z c Let O1 be the camera coordinate system and O1-X1Y1 be the image plane coordinate system. By solving for the spatial circle pose in the camera coordinate system and then obtaining the extrinsic parameter matrix through camera calibration, the pose of the circle in the world coordinate system can be calculated.

[0043] With the camera already calibrated, the center position and circular plane attitude parameters were determined based on the projection ellipse curve of the circular features on the image. The calculation results were ambiguous, and angle constraints were used to eliminate the ambiguity (Wei Zhenzhong, Zhao Zheng, Zhang Guangjun. Angle constraint elimination of ambiguity in spatial circular attitude recognition [J]. Optics and Precision Engineering, 2010, 18(03): 685-691).

[0044] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

[0045] The parts not covered in this invention are the same as or can be implemented using existing technologies.

Claims

1. A method for accurate pose measurement of a large-size rotating ring based on machine vision, characterized in that, Includes the following steps: S1: Use a multi-view vision system to acquire images of a large rotating ring. This multi-view vision system uses four cameras evenly distributed around the rotating ring. S2: Accurately extract key features of the edge arc segment from the acquired image of the rotating ring; S3: Unify the features of multiple circular arc segments to the same coordinate system based on the multi-view camera calibration relationship; S4: Calculate the pose of the rotating ring based on the features of multiple arc segments and the actual size of the rotating ring; Step S2, which involves accurately extracting key features of the edge arc segment from the acquired image of the rotating ring, includes the following steps: S2.1: Image preprocessing of the rotating ring image, including filtering and denoising, image segmentation, and region of interest extraction; S2.2: Accurately detect the projected line segments of the rotating ring arc segment spline; extract key features of the edge arc segment using the Canny adaptive edge detection and RANSAC optimized foreground circular imaging ellipse detection method as a benchmark. First, perform edge detection and use adaptive thresholding for trimming, segmenting the acquired image into foreground and background. The total number of pixels in the image. For the threshold, The proportion of foreground pixels to the total number of pixels. The average gray level, The proportion of background pixels to the total number of pixels. The average gray level is denoted as ; the total gray level of the image is denoted as . The variance between classes is denoted as ; through the The maximum threshold is obtained by optimizing one by one. In Canny, set it to a high threshold; in Canny, set... The threshold is low; in addition, the arc segment is discretized into multiple point sets, and the RANSAC clustering method is used to cluster the point sets to improve the recognition accuracy and robustness. Step S3, which unifies multiple circular arc segment features to the same coordinate system based on the multi-view camera calibration relationship, includes the following steps: S3.1: Use a large calibration board to unify camera coordinates; that is, use a large calibration board to unify the coordinates of all cameras. Each large calibration board contains several smaller calibration boards, and the positional relationships between the smaller calibration boards are known. Each camera can capture images of one of the smaller calibration boards. The internal and external parameters of each camera can be calibrated using these smaller calibration boards, and the coordinates of each camera can be transformed to the coordinate system of each smaller calibration board, thus unifying the coordinates of all cameras. The specific calibration process is as follows: Define: , Representing the camera coordinate system , coordinate system with small calibration plate , The transformation matrix is ​​obtained by solving for the calibration board image captured by the camera; Representing the camera coordinate system , The transformation matrix between them; Represents the coordinate system of the small calibration plate , The transformation matrix between the coordinate systems is obtained during the design and calibration of the fixture; once the coordinate system transformation relationship is determined, the calibration equations can be established. Determine the camera coordinate system , Transformation matrix between Similarly, the transformation matrix between the four cameras is obtained and unified to... ; S3.2: Unify the features of multiple circular arc segments to the same coordinate system according to the calibration parameters of each camera; Step S4, which calculates the pose of the rotating ring based on the features of multiple arc segments and the actual dimensions of the rotating ring, includes the following steps: S4.1: Fit a complete ellipse feature based on multiple arc segment features; S4.2: Calculate the pose of the rotation ring based on the complete elliptical features and the actual dimensions of the rotation ring; after perspective transformation, the spatial circle is imaged as an ellipse on the image plane, which is understood as being intercepted by the elliptical cone formed by the image plane, the spatial circle, and the optical center; by solving the pose of the spatial circle in the camera coordinate system, and then obtaining the extrinsic parameter matrix through camera calibration, the pose of the circle in the world coordinate system can be calculated; with the camera already calibrated, determine the center position and the pose parameters of the circular plane based on the projection elliptical curve of the circle features on the image, and use angular constraints to eliminate ambiguity.

Images