Semi-stereo correction matching method for 3D digital speckle images

Abstract

This invention discloses a semi-stereoscopic correction matching method for three-dimensional digital speckle images, comprising the following steps: S1, acquiring a speckle image of the object to be tested, wherein the speckle image includes an original reference image and an original target image; S2, applying coarse and fine matching to the corrected image and the original image respectively, wherein the corrected image is the reference image and the target image after stereo correction, and the original image is the speckle image acquired in step S1. This invention introduces stereo correction technology to optimize the coarse matching process while avoiding the introduction of additional systematic errors in the fine matching process, effectively improving the efficiency and success rate of stereo matching. Furthermore, it avoids introducing additional systematic errors due to stereo correction. The stereo correction algorithm optimizes the diagonal epipolar search in traditional stereo matching methods into a horizontal line search. The horizontal line search can use the integral image method to accelerate the coarse matching process, which significantly improves the efficiency of coarse matching.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional digital speckle displacement and strain measurement technology, specifically a semi-stereoscopic correction and matching method for three-dimensional digital speckle images. Background Technology

[0002] Digital image correlation (DIC) is a novel non-destructive, non-contact measurement method. Due to its advantages such as relaxed experimental requirements, full-field measurement, high spatial resolution, and high precision, it has wide applications in material property measurement, structural condition monitoring, and biomechanics. DIC is based on photographic recordings of images of the object before and after deformation. Correlation analysis is performed using speckle patterns pre-created on the object's surface to obtain displacement and strain information. DIC includes two-dimensional digital image correlation (2D-DIC) and three-dimensional digital image correlation (3D-DIC). Stereo matching is a crucial step in 3D-DIC; its success and accuracy directly affect the 3D reconstruction effect, and consequently, the final measurement accuracy. Stereo matching in 3D-DIC is generally achieved through coarse and fine matching. First, coarse matching obtains the approximate positions of corresponding points, and then fine matching is performed based on the coarse matching results to improve matching accuracy to meet practical application needs. In classic 3D-DIC stereo matching, coarse matching uses the ZNCC (Zero mean normalized cross correlation) algorithm, while fine matching most often uses the Inverse Compositional Gauss-Newton (IC-GN) algorithm.

[0003] There are two problems in 3D-DIC stereo matching: First, a large overlapping field of view is required during measurement. The binocular camera needs to maintain a certain angle to form a tilted field of view, which causes distortion of the features of the left and right images, thereby reducing the correlation between the left and right images. This may cause stereo matching to fail in the coarse matching stage, and thus cannot perform subsequent fine matching. Second, the ZNCC algorithm has a lot of redundant calculations along the epipolar line, which makes the coarse matching stage too time-consuming, thereby reducing the overall calculation speed of 3D-DIC. Summary of the Invention

[0004] The purpose of this invention is to provide a semi-stereoscopic correction matching method for three-dimensional digital speckle images. It introduces stereoscopic correction technology to optimize the coarse matching process of 3D-DIC, while avoiding the introduction of additional systematic errors in the fine matching process. This can effectively improve the efficiency and success rate of stereoscopic matching, and will not introduce additional systematic errors due to stereoscopic correction, thereby solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a semi-stereoscopic correction and matching method for three-dimensional digital speckle images, comprising the following steps:

[0006] S1. Acquire a speckle image of the object to be tested, wherein the speckle image includes the original reference image and the original target image;

[0007] S2. Perform semi-stereo correction matching on the speckle image, and apply coarse and fine matching to the corrected image and the original image respectively. The corrected image is the reference image and the target image after stereo correction, and the original image is the speckle image obtained in step S1.

[0008] Preferably, in step S2, the semi-stereoscopic correction and matching process is as follows:

[0009] S2.1 Stereo Correction: Perform stereo correction on the original reference image and the original target image to obtain the corrected reference image and the corrected target image;

[0010] S2.2, Reference Sub-region Center Coordinate Correction Transformation: The center coordinates of the reference sub-region in the original reference image are transformed using the correction matrix to obtain the corresponding center coordinates of the reference sub-region in the corrected reference image, and a new reference sub-region is established using these center coordinates;

[0011] S2.3, ZNCC matching: Using the new reference sub-region obtained in S2.2, perform coarse ZNCC matching in the corrected target image;

[0012] S2.4 Perform inverse correction transformation of ZNCC coarse matching results;

[0013] S2.5 IC-GN Matching: Using the initial values ​​of the original target image obtained in S2.4, fine matching is performed using the IC-GN algorithm.

[0014] Preferably, in step S2.1, the Fusiello stereo correction method is used to obtain the correction transformation matrix of the original reference image and the original target image, and the corrected image is calculated. The pixel coordinates of the imaging system before and after correction have the following relationship: Where m ol m or Let m represent the homogeneous coordinates of the original reference image and the original target image before correction, respectively. nl m nr These are the homogeneous coordinates of the corrected reference image and the corrected target image, respectively, T. l T r These are the correction matrices for the original reference image and the original target image, respectively.

[0015] Preferably, in step S2.4, the coarse matching result obtained in the corrected image in S2.3 is transformed into the initial iterative value in the original target image through inverse correction.

[0016] Preferably, in step S2.3, using the new reference sub-region obtained in step S2.2, coarse matching is performed in the stereo-corrected target image using the ZNCC algorithm accelerated by integral image technology; the relevant function expressions of the ZNCC algorithm are as follows:

[0017]

[0018] The pixel gray level and the integral S(x,y) have the following relationship:

[0019] f(x,y)=S(x,y)+S(x-1,y-1)-S(x-1,y)-S(x,y-1),

[0020] Where S(x,y) represents the sum of the gray levels of all pixels in the interval from the first row and first column to the x-th row and y-th column, and f(x,y) is the gray level of a single pixel, y∈[1,n], and y∈Z; similarly, the square of the pixel gray level f can be obtained. 2 (x,y) and integral S 2 The relationship between (x, y):

[0021] f 2 (x,y)=S 2 (x,y)+S 2 (x-1,y-1)-S 2 (x-1,y)-S 2 (x, y-1),

[0022] When it is necessary to calculate the cumulative grayscale sum of pixels within a sub-region centered at (x,y) with a radius of N, the integral method is used:

[0023]

[0024]

[0025] Preferably, in step S2.1, during stereo correction, the expression for the camera projection matrix is: Then the projection matrix P of the left and right cameras before correction o1 ,P o2 Represented as The coordinates of the left and right optical centers c1 and c2 are:

[0026] After correction, the left and right cameras have the same rotation matrix R, and the projection matrix P of the corrected left and right cameras is also the same. n1 and P n2They are respectively represented as Among them, K n1 K n2 This refers to the intrinsic parameters of the calibrated camera. The calibrated rotation matrix R is constructed as follows: the optical center coordinates of the calibrated dual-camera system remain unchanged from the original system. The line connecting c1 and c2 is used as the calibrated X-axis. The cross product of the original left camera's Z-axis and the new X-axis is used as the calibrated Y-axis. The cross product of the calibrated Y-axis and the calibrated X-axis yields the calibrated Z-axis. The rotation matrix is ​​then expressed as... Where r1 = (c1 - c2) / ||c1 - c2||, r2 = k × r1, r3 = r1 × r2, and k represents the Z-axis of the original left camera. The stereo correction matrix is ​​constructed as follows: Where T l T r These represent the transformation matrices between the homogeneous pixel coordinates before and after the left and right camera corrections, respectively.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. The semi-stereo correction matching method of the present invention completes coarse matching in the corrected reference image and the corrected target image, then transforms the center coordinates of the target sub-region obtained by coarse matching into the original target image through inverse correction, and performs fine matching in the uncorrected original target image. Compared with the traditional stereo matching method, it does not change the stereo matching accuracy, thus ensuring the matching accuracy.

[0029] 2. This invention introduces stereo calibration technology to optimize the coarse matching process, while avoiding the introduction of additional systematic errors in the fine matching process. This can effectively improve the efficiency and success rate of stereo matching, and will not introduce additional systematic errors due to stereo calibration.

[0030] 3. The semi-stereo calibration matching method of the present invention uses a stereo calibration algorithm to optimize the oblique epipolar search in the traditional stereo matching method into a horizontal line search. The horizontal line search can use the integral graph method to accelerate the coarse matching process, which significantly improves the efficiency of coarse matching. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method steps of the present invention;

[0032] Figure 2 This is a schematic diagram of the semi-stereoscopic correction method of the present invention;

[0033] Figure 3 This is a comparison diagram of the X-coordinate deviation between semi-stereoscopic correction and stereoscopic correction in an embodiment of the present invention;

[0034] Figure 4 This is a comparison diagram of the Y-coordinate deviation between semi-stereoscopic correction and stereoscopic correction in an embodiment of the present invention;

[0035] Figure 5This is a comparison diagram of the Z-coordinate deviation between semi-stereoscopic correction and stereoscopic correction in an embodiment of the present invention;

[0036] Figure 6 This is a comparison chart of the correlation coefficients between the semi-stereoscopic correction method coarse matching and the traditional coarse matching in an embodiment of the present invention;

[0037] Figure 7 This is a complete view of the steps constructed using the conventional method in this embodiment of the invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," "top / bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0040] Please see Figure 1-2 This invention provides a technical solution: a semi-stereo correction and matching method for three-dimensional digital speckle images, comprising the following steps:

[0041] S1. Acquire a speckle image of the object to be tested, wherein the speckle image includes the original reference image and the original target image;

[0042] S2. Perform semi-stereo correction matching on the speckle image, and apply coarse and fine matching to the corrected image and the original image respectively. The corrected image is the reference image and the target image after stereo correction, and the original image is the speckle image obtained in step S1.

[0043] Specifically: First, stereo calibration: Perform stereo calibration on the original reference image and the original target image to obtain the calibrated reference image and the calibrated target image;

[0044] Then, the reference sub-region center coordinate correction transformation is performed: the reference sub-region center coordinates in the original reference image are transformed using the correction matrix to obtain the corresponding reference sub-region center coordinates in the corrected reference image, and a new reference sub-region is established using these center coordinates.

[0045] Next, ZNCC matching is performed: using the new reference sub-region obtained in S2.2, ZNCC coarse matching is performed in the corrected target image; after completing ZNCC matching, the ZNCC coarse matching result is subjected to inverse correction transformation.

[0046] Finally, IC-GN matching is performed: using the initial values ​​of the original target image obtained in S2.4, fine matching is performed through the IC-GN algorithm.

[0047] In the specific measurement process, the displacement and strain information of the object under test are obtained by analyzing the image information processed in step S2.

[0048] In S2.1, the Fusiello stereo correction method is used to obtain the correction transformation matrix of the original reference image and the original target image, and the corrected image is calculated. The pixel coordinates of the imaging system before and after correction have the following relationship: Where m ol m or Let m represent the homogeneous coordinates of the original reference image and the original target image before correction, respectively. nl m nr These are the homogeneous coordinates of the corrected reference image and the corrected target image, respectively, T. l T r These are the correction matrices for the original reference image and the original target image, respectively.

[0049] In step S2.4, the coarse matching result obtained in the corrected image in S2.3 is transformed into the initial iterative value in the original target image through inverse correction.

[0050] In step S2.3, using the new reference sub-region obtained in step S2.2, coarse matching is performed in the stereo-corrected target image using the ZNCC algorithm accelerated by integral image technology; the relevant function expressions of the ZNCC algorithm are as follows:

[0051]

[0052] The pixel gray level and the integral S(x,y) have the following relationship:

[0053] f(x,y)=S(x,y)+S(x-1,y-1)-S(x-1,y)-S(x,y-1),

[0054] Where S(x,y) represents the sum of the gray levels of all pixels in the interval from the first row and first column to the x-th row and y-th column, and f(x,y) is the gray level of a single pixel, y∈[1,n], and y∈Z; similarly, the square of the pixel gray level f can be obtained. 2 (x,y) and integral S 2 The relationship between (x, y):

[0055] f 2 (x,y)=S 2 (x,y)+S 2 (x-1,y-1)-S 2 (x-1,y)-S 2 (x, y-1)

[0056] When it is necessary to calculate the cumulative grayscale sum of pixels within a sub-region centered at (x,y) with a radius of N, the integral method is used:

[0057]

[0058]

[0059] In step S2.1, during stereo calibration, the expression for the camera projection matrix is: Then the projection matrix P of the left and right cameras before correction o1 ,P o2 Represented as The coordinates of the left and right optical centers c1 and c2 are:

[0060] After correction, the left and right cameras should have the same rotation matrix R, and the corrected projection matrix P of the left and right cameras should also be the same. n1 and P n2 They are respectively represented as Among them, K n1 K n2 These are the intrinsic parameters of the calibrated camera. They can be set as needed, but it is necessary to ensure that the focal length and vertical translation amount of the intrinsic parameters of the left and right cameras are consistent.

[0061] Construction of the corrected rotation matrix R: The optical center coordinates of the corrected dual-camera system remain unchanged from the original system. The line connecting c1 and c2 is used as the corrected X-axis. The cross product of the original left camera's Z-axis and the new X-axis is used as the corrected Y-axis. The cross product of the corrected Y-axis and the corrected X-axis yields the corrected Z-axis. The rotation matrix is ​​then expressed as: Where r1 = (c1 - c2) / ||c1 - c2||, r2 = k × r1, r3 = r1 × r2, and k represents the Z-axis of the original left camera.

[0062] Construction of stereo calibration matrix: T l T r These represent the transformation matrices between the homogeneous pixel coordinates before and after correction for the left and right cameras, respectively. Through this relationship, the pixel coordinates of the original image corresponding to the integer pixel position of the new image can be established, and the corrected image can be calculated through interpolation.

[0063] The superiority of the stereo calibration matching method is verified by measuring the height of a standard step.

[0064] Specifically, measurements were performed using both traditional stereo matching and semi-stereo calibration matching methods. The least squares method was used to reconstruct the three-dimensional image from the stereo matching results. The three-dimensional reconstruction results from the traditional stereo matching method were used as a benchmark, and the offset of the semi-stereo calibration matching method's reconstruction results relative to the benchmark data was compared and analyzed. Furthermore, to further explain why the stereo calibration matching method is not used in three-dimensional speckle displacement and strain measurements—that is, why the fine matching algorithm is also applied to the calibrated image—the three-dimensional reconstruction results from the stereo calibration matching method were also compared with the benchmark data.

[0065] The test computer system was Windows 11, equipped with an Intel(R) Core(TM) i7-11800H CPU @ 2.30GHz, 32GB of RAM, and Matlab 2022a as the test platform. Two HT-SUA520-T monochrome industrial cameras with an effective resolution of 2592×1944 pixels were used in the experiment. To ensure successful matching using all three stereo matching methods, a relatively small solid angle of approximately 19° was used. The calculation parameters selected for this experiment were a sub-region size of 51×51 pixels and a pixel pitch of 8 pixels. Considering that first-order shape functions generally meet the requirements and are easier to converge, the first-order shape function IC-GN algorithm was used for fine matching calculations.

[0066] Using the lowest step as the reference plane, the stepped plane reconstructed by the traditional stereo matching method is as follows: Figure 7 As shown, the 3D reconstruction results of the semi-stereo correction matching method and the stereo correction matching method are compared with the 3D reconstruction results of the traditional stereo matching method. The difference in their X-coordinates is shown in the figure. Figure 3 As shown, the difference in the Y coordinates is as follows Figure 4 As shown, the difference in Z coordinates is as follows Figure 5 As shown.

[0067] Depend on Figure 3 (a) Figure 4 (a) Figure 5 (a) It can be seen that the coordinate deviation reconstructed by the semi-stereo correction matching method is generally small, all around 0.

[0068] Combination Figure 3 (b) Figure 4 (b) Figure 5 (b) It can be seen that the reconstructed coordinate deviation of the stereo calibration matching method in the three directions is significantly greater than that of the semi-stereo calibration matching method.

[0069] Figure 3 (a) Figure 4 (a) Figure 5In (a), the deviation fluctuation range of steps 3 and 4 is slightly larger. This is because speckle quality affects interpolation accuracy. Since it is difficult to ensure the consistency of speckle quality during the fabrication process, the speckle quality on steps 3 and 4 is lower than that on steps 1 and 2. This leads to increased randomness in fine matching, and consequently, a wider range of deviation fluctuation in the reconstruction results. Figure 3 (b) Figure 5 (b) It can be seen that the coordinates reconstructed by the stereo calibration matching method deviate significantly from the reference data in the X and Z directions at the 3rd and 4th steps compared to the 1st and 2nd steps. This is because the stereo calibration method performs secondary interpolation, and the interpolation error caused by the low speckle quality is introduced multiple times, thus increasing the fine matching error and resulting in a larger deviation in the reconstruction result. Further quantitative analysis was conducted on the deviations between the reconstruction results obtained by the semi-stereo calibration matching method and the stereo calibration matching method and the reference data. The maximum and average deviations in the three directions were statistically analyzed, and the results are shown in Table 1.

[0070] Table 1 Error Statistics

[0071]

[0072] As shown in Table 1, the maximum coordinate deviation between the proposed semi-stereo calibration and matching method and the reference data in all three directions is less than 10⁻⁴ mm, and the average deviation is only on the order of 10⁻⁷ mm, which can basically be considered as calculation error. The maximum coordinate deviation between the reconstruction result obtained by the stereo calibration and matching method and the reference data in the X and Y directions is less than 10⁻⁴ mm. -3 The magnitude is on the order of mm, and the maximum coordinate deviation in the Z direction reaches 2 × 10⁻⁶. -2 mm, with an average deviation of 8.4 × 10 -3 mm. Compared to the stereo calibration matching method, the semi-stereo calibration matching method shows significantly smaller maximum and average deviations in reconstruction results compared to the traditional matching method. The difference in deviations in the three directions is because the reconstruction accuracy is not only related to the stereo matching accuracy but also to the structural parameters of the binocular camera system. Due to the influence of structural parameters, the deviation amounts in the three directions will differ when the stereo matching error is reflected in the 3D reconstruction.

[0073] The experiment involved 1557 detection points. The time taken by traditional stereo matching and semi-stereo correction matching to calculate all detection points is shown in Table 2.

[0074] Table 2 Comparison of Calculation Time

[0075]

[0076] As shown in Table 2, thanks to the acceleration effect of the integral image method, the time for semi-stereo calibration matching to complete stereo matching is only half that of the traditional method. Using semi-stereo calibration matching can significantly improve the efficiency of DIC measurement. The essential difference between the semi-stereo calibration matching method and the traditional matching method lies in the different ways of obtaining the initial values ​​for iteration. To discuss the difference in the performance of the two methods in the coarse matching stage, the maximum ZNCC correlation coefficients obtained by the two methods in the coarse matching stage were statistically analyzed. The statistical results are shown below. Figure 6 As shown.

[0077] Depend on Figure 6 It can be seen that during coarse matching using the semi-stereo correction matching method, the maximum ZNCC correlation coefficient between the reference sub-region and the target sub-region is approximately around 0.95, while the minimum value is also higher than 0.8. In contrast, during coarse matching using the traditional stereo matching method, the maximum ZNCC correlation coefficient between the reference sub-region and the target sub-region has a wide range, while the minimum value is only around 0.3. This is because in the semi-stereo correction matching method, both the reference image and the target image undergo stereo correction processing. Image feature distortions caused by camera rotation and different camera intrinsic parameters are corrected to a certain extent, making the reference sub-region and the corresponding target sub-region closer in shape and size. This increases the correlation coefficient between the reference sub-region and the target sub-region, and a larger correlation coefficient means higher reliability of the matching result. The semi-stereo correction matching method obtains more reliable and accurate results in the coarse matching stage, which will be beneficial to the successful matching of the fine matching algorithm and thus improve the success rate of stereo matching.

[0078] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0079] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of semi-stereo rectification matching of three-dimensional digital speckle images, characterized in that, The method comprises the following steps: S1, obtaining a speckle image of an object to be measured, the speckle image comprising an original reference image and an original target image; S2, performing semi-stereo rectification matching on the speckle image, wherein coarse matching and fine matching are performed on the rectified image and the original image, the rectified image being a rectified reference image and a rectified target image, and the original image being the speckle image obtained in step S1; The process of the semi-stereo rectification matching is as follows: S2.1, stereo rectification: performing stereo rectification on the original reference image and the original target image to obtain a rectified reference image and a rectified target image; S2.2, reference sub-region center coordinate rectification transformation: converting the reference sub-region center coordinates in the original reference image to obtain corresponding reference sub-region center coordinates of the rectified reference image by using a rectification matrix, and establishing a new reference sub-region based on the center coordinates; S2.3, ZNCC matching: performing ZNCC coarse matching in the rectified target image by using the new reference sub-region obtained in S2.2; S2.4, performing inverse rectification transformation on the ZNCC coarse matching result; S2.5, IC-GN matching: performing fine matching by using the original target image initial value obtained in S2.4 through the IC-GN algorithm.

2. The method of semi-stereocorrection matching of three-dimensional digital speckle images according to claim 1, characterized in that: In the S2.1, the Fusiello stereo rectification method is adopted to obtain the rectification transformation matrix of the original reference image and the original target image, and to calculate the corrected image. The pixel coordinates of the imaging system before and after the correction have the following relationship: Wherein m ol , m or respectively represent the homogeneous coordinates of the original reference image and the original target image before the correction, m nl , m nr are the homogeneous coordinates of the corrected reference image and the corrected target image respectively, T l , T r are the correction matrices of the original reference image and the original target image respectively.

3. The method of semi-stereocorrection matching of three-dimensional digital speckle images according to claim 2, characterized in that: In step S2.4, the coarse matching result obtained in the rectified image in S2.3 is transformed into the original target image initial value through inverse rectification transformation.

4. The method of semi-stereocorrection matching of three-dimensional digital speckle images according to claim 1, characterized in that: In step S2.3, the new reference sub-region obtained in step S2.2 is used to complete coarse matching in the rectified target image by using the ZNCC algorithm accelerated by the integral graph technology; the expression of the ZNCC algorithm correlation function is as follows: The pixel gray value and the integral S(x, y) have the following relationship: f(x, y) = S(x, y) + S(x-1, y-1) - S(x-1, y) - S(x, y-1), Wherein, S(x, y) represents the accumulated sum of all pixel gray scales in the interval from the first row and the first column to the xth row and the yth column, f(x, y) is a pixel gray scale, y∈[1, n], and y∈Z; similarly, the relationship between the pixel gray scale square f 2 (x, y) and the integral S 2 (x, y) is obtained. f 2 (x,y) = S 2 (x,y) + S 2 (x-1,y-1) - S 2 (x-1,y) - S 2 (x,y-1) When the pixel gray value accumulation sum of a sub-region with (x, y) as the center and N as the radius is needed to be calculated, the integral method is used for calculation:

5. The method of semi-stereo rectification matching of three-dimensional digital speckle images according to claim 1, characterized in that: The expression of the camera projection matrix in the step S2.1 of the stereoscopic correction process is Then the left and right camera projection matrixes P o1 ,P o2 are expressed as The coordinates of the left and right optical centers c1, c2 are The left and right cameras have the same rotation matrix R after correction, and the projection matrix P of the left and right cameras after correction n1 and P n2 are respectively expressed as wherein K n1 , K n2 is the intrinsic parameter of the camera after correction; The corrected rotation matrix R is constructed: the optical center coordinates of the corrected two-camera system are invariant to the original system, the line c1c2 is taken as the X axis of the corrected system, the cross product of the Z axis of the original left camera and the new X axis is taken as the Y axis of the corrected system, the cross product of the corrected Y axis and the corrected X axis is taken as the Z axis of the corrected system, and the rotation matrix is represented as wherein r1=(c1-c2) / ||c1-c2||, r2=kxr1, r3=r1xr2, k represents the Z axis of the original left camera, and the stereo correction matrix is constructed as: wherein T l , T r respectively represent the conversion matrixes between the pixel homogeneous coordinates before and after the correction of the left and right cameras.

Images