Binocular stereoscopic vision system and structured light 3D scanner composite calibration method

Abstract

The invention provides a binocular stereoscopic vision system and structured light 3D scanner composite calibration method, and the method employs a digital coding point and checkerboard composite calibration technology. Only four digital coding points and a common checkerboard are needed to realize conversion matrix solution from a coding point coordinate system constructed by the digital coding points to a structured light 3D scanner measurement coordinate system, and high-precision dynamic calibration of a binocular vision system and a structured light scanner is realized. The problem of adaptability of a traditional calibration method in a complex industrial environment is solved, the calibration precision and stability of a composite measurement system are remarkably improved, meanwhile, the calibration method has the advantages of being simple in calibration process, low in system cost and the like, the calibration precision and stability are improved, a brand new technical solution is provided for three-dimensional measurement of an industrial site, and the method is worthy of popularization and application. The method is suitable for a dynamic measurement scene in a complex industrial environment.

Description

Technical Field

[0001] This invention belongs to the field of three-dimensional vision measurement and multi-sensor fusion technology, and particularly relates to a composite calibration method of binocular stereo vision system and structured light 3D scanner. Background Technology

[0002] With the increasing demands for 3D measurement accuracy in modern manufacturing, structured light 3D scanning technology, with its non-contact measurement advantages, is playing an increasingly important role in industrial inspection. However, this technology still faces pressing technical bottlenecks in engineering applications. Due to the limited field of view of a single structured light 3D scanner, multi-view scanning and data stitching are necessary to complete the overall measurement when measuring large workpieces. Achieving accurate registration of measurement data from different perspectives is a key challenge in this process.

[0003] Currently, the common industrial solution is to establish a global coordinate system using laser trackers. While this method solves the problem of large-scale measurement to some extent, it has revealed several limitations in practical applications. First, the purchase and maintenance costs of the entire laser tracking system are high, placing a heavy economic burden on users. Second, these high-precision optical devices have stringent environmental requirements; in industrial environments with vibration or temperature drift, measurement stability is difficult to guarantee. More importantly, existing calibration methods rely on dedicated target devices, which not only increases operational complexity but also introduces additional sources of error. These factors restrict the widespread application of this technology in industrial settings.

[0004] Limitations of Existing Calibration Methods: Current mainstream calibration schemes mainly fall into two categories: one is based on dedicated target sets, and the other is based on replacement measurements. The former, for example, is illustrated in Chinese patent CN 110440692 B, which discloses a composite measurement and calibration method using a laser tracker and a structured light 3D scanner. This method requires four precisely calibrated laser target balls, which are not only expensive but also require regular calibration to ensure accuracy. In practice, the installation and positioning of the target set is time-consuming and easily affected by environmental interference. The latter, while reducing costs to some extent, inevitably introduces systematic errors due to the need to replace target balls between different measurement devices. Furthermore, the limited field of view of the structured light scanner restricts the number of target balls that can be identified in a single measurement, further impacting calibration accuracy and efficiency. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a composite calibration method for a binocular stereo vision system and a structured light 3D scanner. By constructing a multi-level calibration system, high-precision collaborative calibration of the binocular vision system and the structured light scanner can be achieved with only four digital coding points and a standard checkerboard grid, significantly simplifying the calibration process and greatly reducing equipment costs.

[0006] A method for calibrating a binocular stereo vision system combined with a structured light 3D scanner includes the following steps:

[0007] The structured light 3D scanner, binocular stereo vision system, and checkerboard are deployed in such a way that their relative positions satisfy the following conditions: both the structured light 3D scanner and the binocular stereo vision system can capture the complete checkerboard surface, and the structured light 3D scanner falls within the field of view of the binocular stereo vision system.

[0008] A structured light 3D scanner and a binocular stereo vision system were used to simultaneously capture chessboard images at different poses. The measurement coordinate system O from the structured light 3D scanner was then determined based on these chessboard images. sc -x sc y sc z sc To global coordinate system O tc -x tc y tc z tc Transformation matrix T sc2tc ;

[0009] A binocular stereo vision system is used to capture images of each coded point arranged on a structured light 3D scanner. The coordinate system O of the coded points is then determined based on the images of each coded point. b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc ;

[0010] According to matrix chain operations T sc2b =T -1 b2tc ·T sc2tc The measurement coordinate system O is obtained. sc -x sc y sc z sc To the coordinate system O of the encoding point b -x b y b z b Transformation matrix T sc2b The calibration is now complete.

[0011] Furthermore, the measurement coordinate system O sc -x sc y sc z sc To global coordinate system O tc -x tc y tc z tc Transformation matrix T sc2tc The solution method is as follows:

[0012] By manipulating a structured light 3D scanner and a binocular stereo vision system, the complete surface of the chessboard grid, which moves 19 to 25 poses within a common measurement range, is captured to obtain chessboard grid images under different poses.

[0013] Left and right images containing a checkerboard pattern were acquired using a structured light 3D scanner. These images were then processed using the Zhang Zhengyou calibration method to obtain the corner points of the checkerboard pattern at position O under different poses. sc -x sc y sc z sc The coordinates below (X) sc1 ,Y sc1 Z sc1 ), (X sc2 ,Y sc2 Z sc2 )…(X scn ,Y scn Z scn ), where n ranges from 10 to 20;

[0014] Left and right images containing a checkerboard pattern were captured using a binocular stereo vision system. These images were then processed using the Zhang Zhengyou calibration method to obtain the corresponding checkerboard corner points at position O under different poses. tc -x tc y tc z tc The coordinates below (X) tc1 ,Y tc1 Z tc1 ), (X tc2 ,Y tc2 Z tc2 )……(X tcn ,Y tcn Z tcn );

[0015] Let O sc -x sc y sc z sc The set of corner points of the chessboard is matrix P. sc O tc -x tc y tc ztc The set of corner points of the chessboard is matrix P. tc From O sc -x sc y sc z sc To O tc -x tc y tc z tc The transformation matrix is ​​T sc2tc ,but:

[0016]

[0017] Where, p tc1 p tc2 ...p tcn O tc -x tc y tc z tc The matrix P below tc The coordinate vector of p; sc1 p sc2 ...p scn O sc -x sc y sc z sc The matrix P below sc The coordinate vector of P, and P sc With P tc The relationships that are satisfied are:

[0018] T sc2tc ·P sc =P tc

[0019] The objective function constructed based on Protodyakonov's analysis theory using the SVD method is as follows:

[0020]

[0021] In the formula, s is the scale factor; R represents the rotation matrix, and t represents the translation matrix;

[0022] Find the centroids of the two sets of corner points on the chessboard. Then for matrix P tc P sc Decentralization results in:

[0023]

[0024] Where G is matrix P sc The result after decentralization, H is matrix P tc The decentralized result, ones(1,n), represents a 1×n matrix with all elements being 1;

[0025] G·H T Performing SVD decomposition, we obtain the rotation matrix R and the translation matrix t as follows:

[0026]

[0027] Where U and V are both auxiliary matrices in the SVD decomposition, and satisfy U T U = I, V T V = I, where I is the identity matrix;

[0028] Based on the calculated rotation matrix R and translation matrix t, calculate from O sc -x sc y sc z sc To O tc -x tc y tc z tc Transformation matrix T sc2tc .

[0029] Furthermore, the structured light 3D scanner has four coding points, two of which are located on the body of the structured light 3D scanner and two of which are located on the fixture of the structured light 3D scanner.

[0030] Furthermore, the coordinate system O of the coding point b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc The solution method is as follows:

[0031] The YOLOv8 convolutional neural network is used to identify the type information of digit coding points and the location of the digit recognition box in the coding point image. Non-target region masking is performed on the non-boxed areas in the coding point image. The processed image is then used with Canny edge detection and morphological fitting algorithm to obtain the pixel coordinates of the four coding point localization centers C1, C2, C3 and C4 as (X1,Y1,Z1), (X2,Y2,Z2), (X3,Y3,Z3) and (X4,Y4,Z4) respectively.

[0032] Take the first encoded point as the origin of the coordinate system. The vectors of C1 and C2 After normalization, it is used as the x-axis of the coordinate system, and the x-axis vector is represented as...

[0033] Find the vector and The vector product of the vectors is normalized to obtain the z-axis of the coordinate system, where the z-axis vector is...

[0034] Calculate the y-axis vector based on the z-axis vector and the x-axis vector.

[0035] Based on the x-axis vector, y-axis vector, and z-axis vector, the coordinate system O of the encoding point is obtained. b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc :

[0036]

[0037] Furthermore, the center of any coding point is a solid black circle used as a positioning circle, the outside of the positioning circle is a white ring used as an auxiliary ring, the auxiliary ring falls into the digital coding area, and the rest is a gray background, the center of the gray background, the center of the positioning circle and the center of the auxiliary ring coincide with each other.

[0038] Furthermore, any three coding points cannot be located on the same straight line, and the triangle formed by the three coding points does not contain any obtuse angles greater than 120 degrees.

[0039] Furthermore, the checkerboard pattern has an external dimension of 150×150mm, with a 12×9 square black and white checkerboard pattern drawn on the surface. Each square has a side length of 10mm and a machining accuracy of ±0.01mm.

[0040] Furthermore, the spacing between the structured light 3D scanner and the binocular stereo vision system is 1.2 meters.

[0041] Beneficial effects:

[0042] This invention provides a composite calibration method for a binocular stereo vision system and a structured light 3D scanner. Employing a composite calibration technique combining digital coding points and a checkerboard grid, it constructs a multi-level calibration system. Only four digital coding points and a standard checkerboard grid are needed to solve the transformation matrix from the coordinate system of the digital coding points to the measurement coordinate system of the structured light 3D scanner. This achieves high-precision dynamic calibration of the binocular vision system and the structured light scanner. It not only solves the adaptability problem of traditional calibration methods in complex industrial environments but also significantly improves the calibration accuracy and stability of the composite measurement system. Furthermore, it offers advantages such as a simple calibration process and low system cost, providing a novel technical solution for 3D measurement in industrial settings and is applicable to dynamic measurement scenarios in complex industrial environments. Attached Figure Description

[0043] Figure 1 A flowchart of the composite calibration method for binocular stereo vision and structured light 3D scanner based on digital coding points provided by the present invention;

[0044] Figure 2 A schematic diagram of the composite calibration method of binocular stereo vision and structured light 3D scanner provided by the present invention;

[0045] Figure 3 This is a schematic diagram of coordinate system transformation for the binocular stereo vision and structured light 3D scanner composite calibration method provided by the present invention; wherein, O tc -x tc y tc z tc —Global coordinate system of binocular stereo vision system, O sc -x sc y sc z sc —Structured light 3D scanner measurement coordinate system, O b -x b y b z b —The coordinate system of the coding points constructed by the digital coding points, T b2tc — Coordinate system from O b -x b y b z b To O tc -x tc y tc z tc The transformation matrix, T sc2tc — Coordinate system from O sc -x sc y sc z sc To O tc -x tc y tc ztc The transformation matrix, Tsc2b—coordinate system from O sc -x sc y sc z sc To O b -x b y b z b The transformation matrix, i.e., the calibration matrix;

[0046] Figure 4 The digital code point numbered 4 is provided by this invention;

[0047] 1- Structured light 3D scanner, 2- Fixture, 3- Digital coding points, 4- Binocular stereo vision system, 5- Checkerboard pattern. Detailed Implementation

[0048] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0049] This invention innovatively proposes a solution based on a composite calibration of digital coding points and a checkerboard pattern, such as... Figure 1 As shown, the specific steps include:

[0050] S1: Deploy a structured light 3D scanner, a binocular stereo vision system, and a checkerboard pattern. The positional relationship between the three satisfies the following conditions: both the structured light 3D scanner and the binocular stereo vision system can capture the complete checkerboard pattern surface, and the structured light 3D scanner falls within the field of view of the binocular stereo vision system.

[0051] It should be noted that the binocular stereo vision system used in this embodiment consists of two Daheng MER-500-79U3 industrial imaging cameras, and the structured light 3D scanner is a VMD02-40703D scanner from Shanghai Tuyang Information Technology Co., Ltd. The fixture is made of aluminum alloy and has threaded holes. The checkerboard calibration plate is made of alumina, with dimensions of 150×150mm, and a 12×9 square black and white checkerboard pattern drawn on its surface. Each square has a side length of 10mm, and the machining accuracy is ±0.01mm. Figure 4 As shown, the digit-coded markers are square in shape with a gray background. The center is a solid black circle, used as a positioning circle. Outside the positioning circle is a white annulus, serving as an auxiliary circle to separate the positioning circle from the background and the digits, allowing for more accurate extraction of the positioning circle's outline and thus more precise center identification. The center of the square background, the center of the positioning circle, and the center of the auxiliary annulus coincide with each other.

[0052] like Figure 2 As shown, the specific deployment method of the structured light 3D scanner, binocular stereo vision system, and checkerboard pattern is as follows:

[0053] The structured light 3D scanner 1 is mounted on the robotic arm fixture 2 and secured with screws. Four digital coding points 3 are then tightly attached to the scanner and fixture 2, ensuring proper distribution. The first three points (P0, P1, P2) cannot be on the same straight line, and the triangle formed by the first three points should not contain an obtuse angle greater than 120 degrees. The distance between adjacent digital coding points should be at least 3 cm. The fourth point (P3) is used for verification. The structured light 3D scanner 1 and the binocular stereo vision system 4 are then placed and fixed in suitable positions, with a distance of approximately 1.2 meters between them. The binocular stereo vision system 4 should be able to clearly capture the four digital coding points 3. During calibration, the structured light 3D scanner 1 and the binocular stereo vision system 4 should maintain their poses and be unaffected by vibration. Finally, a checkerboard 5 is placed as a common calibration plate in the common measurement area of ​​the structured light 3D scanner 1 and the binocular stereo vision system 4. The position of the common checkerboard 5 calibration plate should be appropriate so that both the structured light 3D scanner 1 and the binocular stereo vision system 4 can capture a complete and clear image of the checkerboard 5 surface. It should be ensured that the vibration generated when moving the checkerboard will not affect the structured light 3D scanner 1 and the binocular stereo vision system 4. After checking the field of view, the structured light scanner and the binocular vision system are turned on and preheated.

[0054] S2: Simultaneously capture chessboard images at different poses using a structured light 3D scanner and a binocular stereo vision system. Solve the measurement coordinate system O from the structured light 3D scanner based on the chessboard images. sc -x sc y sc z sc To global coordinate system O tc -x tc y tc z tc Transformation matrix T sc2tc ;

[0055] Specifically, the transformation matrix T sc2tc The solution method is as follows:

[0056] By manipulating a structured light 3D scanner and a binocular stereo vision system, the complete surface of the chessboard grid, which moves 19 to 25 poses within a common measurement range, is captured to obtain chessboard grid images under different poses.

[0057] Left and right images containing a checkerboard pattern were acquired using a structured light 3D scanner. These images were then processed using the Zhang Zhengyou calibration method to obtain the corner points of the checkerboard pattern at position O under different poses. sc -x sc y sc z sc The coordinates below (X) sc1 ,Ysc1 Z sc1 ), (X sc2 ,Y sc2 Z sc2 )…(X scn ,Y scn Z scn ), where n ranges from 10 to 20;

[0058] Left and right images containing a checkerboard pattern were captured using a binocular stereo vision system. These images were then processed using the Zhang Zhengyou calibration method to obtain the corresponding checkerboard corner points at position O under different poses. tc -x tc y tc z tc The coordinates below (X) tc1 ,Y tc1 Z tc1 ), (X tc2 ,Y tc2 Z tc2 )……(X tcn ,Y tcn Z tcn );

[0059] Let O sc -x sc y sc z sc The set of corner points of the chessboard is matrix P. sc O tc -x tc y tc z tc The set of corner points of the chessboard is matrix P. tc From O sc -x sc y sc z sc To O tc -x tc y tc z tc The transformation matrix is ​​T sc2tc ,but:

[0060]

[0061] Where, p tc1 p tc2 ...p tcn O tc -x tc y tc z tc The matrix P below tc The coordinate vector of p; sc1 p sc2 ...p scnO sc -x sc y sc z sc The matrix P below sc The coordinate vector of P, and P sc With P tc The relationships that are satisfied are:

[0062] T sc2tc ·P sc =P tc

[0063] It should be noted that, to solve T sc2tc This requires Protodyakonov analysis theory based on the SVD method; obviously, T sc2tc The optimal solution should satisfy ||T sc2tc ·P sc -P tc The minimum, i.e., T sc2tc =argmin(||T sc2tc ·P sc -P tc || F Based on this, the objective function of this invention, constructed using Protodyakonov's analysis theory based on the SVD method, is as follows:

[0064]

[0065]

[0066] In the formula, s is the scale factor; R represents the rotation matrix, and t represents the translation matrix;

[0067] Find the centroids of the two sets of corner points on the chessboard. Then for matrix P tc P sc Decentralization results in:

[0068]

[0069] Where G is matrix P sc The result after decentralization, H is matrix P tc The decentralized result, ones(1,n), represents a 1×n matrix with all elements being 1;

[0070] G·H T Performing SVD decomposition, we obtain the rotation matrix R and the translation matrix t as follows:

[0071]

[0072] Where U and V are both auxiliary matrices in the SVD decomposition, and satisfy U T U = I, VT V = I, where I is the identity matrix;

[0073] Based on the calculated rotation matrix R and translation matrix t, calculate from O sc -x sc y sc z sc To O tc -x tc y tc z tc Transformation matrix T sc2tc .

[0074] Therefore, it can be seen that the present invention obtains the transformation matrix T sc2tc The process can be summarized as follows:

[0075] The structured light 3D scanner 1 and the binocular stereo vision system 4 are used to simultaneously capture images of the complete surface of the common chessboard 5, which moves 19 to 25 poses within the common measurement range. Left and right images containing the chessboard 5 are captured using the structured light 3D scanner (1). Using the Zhang Zhengyou calibration method, the corner points of the chessboard at O ​​can be determined. sc -x sc y sc z sc The coordinates under the given position can be calculated. Similarly, the coordinates of the chessboard corner points under different poses can be calculated at position O. tc -x tc y tc z tc The coordinates are determined. Then, the point sets in the two coordinate systems (each taking 10 corresponding corner points) are used to solve for the coordinates measured from the structured light 3D scanner using the weighted SVD Protodyakonov analysis algorithm. sc -x sc y sc z sc To the global coordinate system O of the binocular stereo vision system tc -x tc y tc z tc The transformation matrix is ​​T sc2tc The measured data and calculation results are shown in Tables 1 and 2.

[0076] Table 1

[0077]

[0078] Table 2

[0079]

[0080] S3: Use a binocular stereo vision system to capture images of each coded point arranged on the structured light 3D scanner, and solve for the coordinate system O of the coded points based on the images of each coded point. b -xb y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc ;

[0081] Specifically, the transformation matrix T b2tc The solution method is as follows:

[0082] The binocular stereo vision system 4 is used to clearly capture images of four specific digital coding points 3 arranged on the structured light 3D scanner 1 and its fixture 2. The Yolov8 convolutional neural network is used to identify the type information of the digital coding points in the coding point images and the location of the digital recognition boxes. Non-target area masking is performed on the non-boxed areas in the coding point images. The processed images are then used with Canny edge detection and morphological fitting algorithms to obtain the pixel coordinates of the four coding point positioning centers C1, C2, C3 and C4 as (X1,Y1,Z1), (X2,Y2,Z2), (X3,Y3,Z3), and (X4,Y4,Z4) respectively.

[0083] Take the first encoded point as the origin of the coordinate system. The vectors of C1 and C2 After normalization, it is used as the x-axis of the coordinate system, and the x-axis vector is represented as...

[0084] Find the vector and The vector product of the vectors is normalized to obtain the z-axis of the coordinate system, where the z-axis vector is...

[0085] Calculate the y-axis vector based on the z-axis vector and the x-axis vector.

[0086] Based on the x-axis vector, y-axis vector, and z-axis vector, the coordinate system O of the encoding point is obtained. b -x b y b z b To global coordinate system O tc -x tc y tc z tc The 4×4 transformation matrix T b2tc :

[0087]

[0088] Therefore, it can be seen that the present invention obtains the transformation matrix T b2tc The process can be summarized as follows:

[0089] The binocular stereo vision system 4 is used to capture images of four specific digital coding points 3 arranged on the structured light 3D scanner 1 and its fixture 2. A Yolov8 convolutional neural network is used for recognition. The pixel coordinates of the positioning centers of the four coding points 3 are obtained using the Canny edge detection algorithm. Combined with the intrinsic and extrinsic parameters of the binocular stereo vision system 4, the center positioning coordinates of the four digital coding points 3 are calculated. Based on the spatial coordinates of the first three points, a transition coordinate system is constructed, and the transition coordinate system O is solved. b -x b y b z b Attached to global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc The measured data and calculation results are shown in Tables 3 and 4.

[0090] Table 3

[0091]

[0092] Table 4

[0093]

[0094] S4: Based on matrix chain operations T sc2b =T -1 b2tc ·T sc2tc The measurement coordinate system O is obtained. sc -x sc y sc z sc To the coordinate system O of the encoding point b -x b y b z b Transformation matrix T sc2b The calibration was completed, and the results are shown in Table 5.

[0095] Table 5

[0096]

[0097] In summary, the composite calibration method of binocular stereo vision and structured light 3D scanner based on digitally encoded points provided by this invention can be summarized as follows:

[0098] A composite measurement and calibration method was employed, using a structured light 3D scanner, a binocular stereo vision system, digitally coded points, and a checkerboard pattern. The coordinate system of the binocular stereo vision system was set as the global coordinate system, the digitally coded points as the transition coordinate system, and the structured light 3D scanner as the measurement coordinate system. First, the checkerboard pattern was changed 19–25 poses within a common measurement range. The spatial coordinates of the corner points of the checkerboard pattern acquired by the structured light 3D scanner and the binocular stereo vision system were solved using the Zhang Zhengyou calibration method, obtaining the coordinate matrices of the corner points in different coordinate systems. Finally, a Protodyakonov analysis algorithm based on weighted SVD was used to calculate the transformation matrix from the measurement coordinate system to the global coordinate system. Next, the binocular stereo vision system can acquire images of specific digitally encoded points arranged on a structured light 3D scanner and its fixture. It uses a Yolov8 convolutional neural network to identify the type information of the digitally encoded points and the location of the recognition boxes in the images. Non-recognition box locations in the images are masked. The processed images are then analyzed using the Canny edge detection algorithm to obtain the pixel coordinates of the center of each encoded point. Based on the intrinsic and extrinsic parameters of the binocular system's camera obtained through Zhang Zhengyou's calibration method, the center coordinates of the digitally encoded points are calculated. A transition coordinate system is constructed using feature points, and the transformation matrix from the transition coordinate system to the global coordinate system is obtained. Finally, through matrix operations, the transformation matrix from the measurement coordinate system to the transition coordinate system of the structured light 3D scanner is calculated, completing the calibration.

[0099] Therefore, this invention innovatively employs a composite calibration technology combining digital coding points and a checkerboard grid. By constructing a multi-level calibration system, high-precision collaborative calibration of a binocular vision system and a structured light scanner can be achieved with only four digital coding points and a standard checkerboard grid. This method significantly simplifies the calibration process, greatly reduces equipment costs, and simultaneously improves calibration accuracy and stability, making it suitable for dynamic measurement scenarios in complex industrial environments.

[0100] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for calibrating a binocular stereo vision system combined with a structured light 3D scanner, characterized in that, Includes the following steps: The structured light 3D scanner, binocular stereo vision system, and checkerboard are deployed in such a way that their relative positions satisfy the following conditions: both the structured light 3D scanner and the binocular stereo vision system can capture the complete checkerboard surface, and the structured light 3D scanner falls within the field of view of the binocular stereo vision system. A structured light 3D scanner and a binocular stereo vision system were used to simultaneously capture chessboard images at different poses. The measurement coordinate system O from the structured light 3D scanner was then determined based on these chessboard images. sc -x sc y sc z sc To global coordinate system O tc -x tc y tc z tc Transformation matrix T sc2tc ; A binocular stereo vision system is used to capture images of each coded point arranged on a structured light 3D scanner. The coordinate system O of the coded points is then determined based on the images of each coded point. b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc ; According to matrix chain operations T sc2b =T -1 b2tc ·T sc2tc The measurement coordinate system O is obtained. sc -x sc y sc z sc To the coordinate system O of the encoding point b -x b y b z b Transformation matrix T sc2b The calibration is now complete.

2. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 1, characterized in that, Measurement coordinate system O sc -x sc y sc z sc To global coordinate system O tc -x tc y tc z tc Transformation matrix T sc2tc The solution method is as follows: By manipulating a structured light 3D scanner and a binocular stereo vision system, the complete surface of the chessboard grid, which moves 19 to 25 poses within a common measurement range, is captured to obtain chessboard grid images under different poses. Left and right images containing a checkerboard pattern were acquired using a structured light 3D scanner. These images were then processed using the Zhang Zhengyou calibration method to obtain the corner points of the checkerboard pattern at position O under different poses. sc -x sc y sc z sc The coordinates below (X) sc1 ,Y sc1 Z sc1 ), (X sc2 ,Y sc2 Z sc2 )…(X scn ,Y scn Z scn ), where n ranges from 10 to 20; Left and right images containing a checkerboard pattern were captured using a binocular stereo vision system. These images were then processed using the Zhang Zhengyou calibration method to obtain the corresponding checkerboard corner points at position O under different poses. tc -x tc y tc z tc The coordinates below (X) tc1 ,Y tc1 Z tc1 ), (X tc2 ,Y tc2 Z tc2 )……(X tcn ,Y tcn Z tcn ); Let O sc -x sc y sc z sc The set of corner points of the chessboard is matrix P. sc O tc -x tc y tc z tc The set of corner points of the chessboard is matrix P. tc From O sc -x sc y sc z sc To O tc -x tc y tc z tc The transformation matrix is ​​T sc2tc ,but: Where, p tc1 p tc2 ...p tcn O tc -x tc y tc z tc The matrix P below tc The coordinate vector of p; sc1 p sc2 ...p scn O sc -x sc y sc z sc The matrix P below sc The coordinate vector of P, and P sc With P tc The relationships that are satisfied are: T sc2tc ·P sc =P tc The objective function constructed based on Protodyakonov's analysis theory using the SVD method is as follows: In the formula, s is the scale factor; R represents the rotation matrix, and t represents the translation matrix; Find the centroids of the two sets of corner points on the chessboard. Then for matrix P tc P sc Decentralization results in: Where G is matrix P sc The result after decentralization, H is matrix P tc The decentralized result, ones(1,n), represents a 1×n matrix with all elements being 1; G·H T Performing SVD decomposition, we obtain the rotation matrix R and the translation matrix t as follows: Where U and V are both auxiliary matrices in the SVD decomposition, and satisfy U T U = I, V T V = I, where I is the identity matrix; Based on the calculated rotation matrix R and translation matrix t, calculate from O sc -x sc y sc z sc To O tc -x tc y tc z tc Transformation matrix T sc2tc .

3. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 1, characterized in that, The structured light 3D scanner has four coding points, two of which are located on the scanner body and two on the scanner fixture.

4. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 3, characterized in that, Coordinate system O of encoding points b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc The solution method is as follows: The YOLOv8 convolutional neural network is used to identify the type information of digit coding points and the location of the digit recognition box in the coding point image. Non-target region masking is performed on the non-boxed areas in the coding point image. The processed image is then used with Canny edge detection and morphological fitting algorithm to obtain the pixel coordinates of the four coding point localization centers C1, C2, C3 and C4 as (X1,Y1,Z1), (X2,Y2,Z2), (X3,Y3,Z3) and (X4,Y4,Z4) respectively. Take the first encoded point as the origin of the coordinate system. The vectors of C1 and C2 After normalization, it is used as the x-axis of the coordinate system, and the x-axis vector is represented as... Find the vector and The vector product of the vectors is normalized to obtain the z-axis of the coordinate system, where the z-axis vector is... Calculate the y-axis vector based on the z-axis vector and the x-axis vector. Based on the x-axis vector, y-axis vector, and z-axis vector, the coordinate system O of the encoding point is obtained. b -x b y b z b To global coordinate system O tc -x tc y tc z tc Transformation matrix T b2tc :

5. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 3, characterized in that, The center of any coding point is a solid black circle used as a positioning circle, and the outside of the positioning circle is a white ring used as an auxiliary ring. The auxiliary ring falls into the digital coding area, and the rest is a gray background. The center of the gray background, the center of the positioning circle, and the center of the auxiliary ring coincide with each other.

6. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 3, characterized in that, No three coding points can be located on the same straight line, and the triangle formed by the three coding points does not contain any obtuse angle greater than 120 degrees.

7. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 1, characterized in that, The checkerboard pattern has external dimensions of 150×150mm, with a 12×9 square black and white checkerboard pattern drawn on the surface. Each square has a side length of 10mm and a machining accuracy of ±0.01mm.

8. The method for calibrating a binocular stereo vision system combined with a structured light 3D scanner as described in claim 1, characterized in that, The spacing between the structured light 3D scanner and the binocular stereo vision system is 1.2 meters.

Images