A research method and system for three-dimensional measurement of cracks based on the fusion of multiple coordinate systems

Abstract

The present invention provides a research method and system for three-way measurement of cracks based on the fusion of multiple coordinate systems, which relates to the technical field of earth-rock dam monitoring. The method includes: taking multi-angle images of concentric circle targets through a monocular camera, extracting feature points after preprocessing and correcting the projection distortion of non-parallel shooting; establishing the mapping relationship between pixels and the three-dimensional coordinate system and optimizing the camera pose parameters; reconstructing the three-dimensional coordinate point cloud based on multi-view data, screening out outliers and fitting the plane to optimize the data; registering the three-dimensional point cloud to the world coordinate system through a coordinate system conversion algorithm, and calculating the three-way displacement of the crack by comparing the two measurement results. The present invention uses the concentric circle target to constrain the feature accuracy, combines projection correction and plane fitting to compensate for the lack of monocular vision depth, realizes high-precision displacement detection, and meets the requirements of engineering monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of earth-rock dam monitoring, and particularly to a research method and system for three-dimensional measurement of cracks based on the fusion of multiple coordinate systems. Background Art

[0002] As a main evaluation index of structural safety, information collection of cracks often adopts manual collection, and the accuracy of measurement results is difficult to guarantee, and the time and labor costs are high. With the development of computer technology, three-dimensional (3D) measurement technology has begun to be applied to the long-term detection of cracks to judge the development status of cracks. Zhang Haoyu et al. established a three-dimensional model through deep learning to realize three-dimensional measurement of cracks; to improve efficiency, Xue Zhilin proposed a stereo vision three-dimensional displacement measurement method based on a dual coordinate system, and Hu Henan integrated motion features and depth information by extracting time-domain features of long and short ranges, and used deep learning to obtain the three-dimensional displacement of an object; Huang Youchao used a binocular camera to establish an equivalent model and constructed a crack measurement system. These methods either have requirements for the sample size and long operation time, or have requirements for equipment and shooting postures, and cannot be well adapted to on-site crack measurement. Summary of the Invention

[0003] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a research method and system for three-dimensional measurement of cracks based on the fusion of multiple coordinate systems. By specifying a world coordinate system, it aims to establish the connection between it and the pixel coordinate system, solve the relative pose between cameras, and perform three-dimensional reconstruction to avoid the problem of missing depth information in monocular vision. By establishing an equivalent model of three-dimensional displacement of cracks, observe the change of the secondary plate coordinates in the world coordinate system to realize three-dimensional displacement measurement of cracks.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A research method for three-dimensional measurement of cracks based on the fusion of multiple coordinate systems, comprising:

[0006] Taking two target layout photos from different positions by the same camera to obtain initial images;

[0007] Performing binarization and filtering processing on the initial images to obtain preprocessed images;

[0008] Extracting concentric circle edge information in the preprocessed images, and obtaining an initial feature point set through a density-based spatial clustering algorithm;

[0009] Performing distortion correction on the initial feature point set based on the invariance of cross-ratio to eliminate the elliptical projection error caused by non-parallel shooting, and obtaining a corrected feature point set;

[0010] Calculate the rotation matrix and translation vector for the two shootings initially, and optimize the reprojection error to below the threshold of 0.1 in combination with the Levenberg-Marquardt method, and output the projection matrix;

[0011] Based on the internal parameter matrix of camera calibration and the said projection matrix, solve the initial three-dimensional coordinate point set in the camera coordinate system by using the least squares method;

[0012] Pre-screen the three-dimensional point cloud data in the said initial three-dimensional coordinate point set through the Random Sample Consensus algorithm, remove the outliers in combination with the principal component analysis, iteratively fit the plane and project, and generate the optimized three-dimensional coordinate point set;

[0013] Adopt the Iterative Closest Point algorithm optimized based on the mean square error, perform point cloud registration according to the said optimized three-dimensional coordinate point set, convert the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three-way displacement change of the crack through the difference between the two measurement results.

[0014] Preferably, the target is a preset concentric circle target, and the center point set of the concentric circles is used as the feature point set.

[0015] Preferably, calculating the rotation matrix and translation vector for the two shootings initially, and optimizing the reprojection error to below the threshold of 0.1 in combination with the Levenberg-Marquardt method, and outputting the projection matrix, includes:

[0016] Establish the mapping relationship between the world coordinate system and the pixel coordinate system through the EPnP algorithm, and solve the rotation matrix R l 、 R r and the translation vector t l 、 t r ;

[0017] Substitute R l 、 R r 、 t l 、 t r into the formula to associate the left and right camera coordinate systems, and obtain the relative rotation matrix and translation vector R and the translation vector t;

[0018] Iteratively optimize R and t through the Levenberg-Marquardt method until the reprojection error is less than 0.1, and output the projection matrix M.

[0019] Preferably, the expression of the mapping relationship between the world coordinate system and the pixel coordinate system is:

[0020]

[0021] Among them, , and are the coordinate values of the , , -axis, X -axis, and Y -axis of point D( Z ) in the world coordinate system, u and v are the coordinate values of the u , v -axis and X -axis of point P( Y ) in the pixel coordinate system, Z c is the coordinate value of the Z -axis in the camera coordinate system, K is the internal parameter matrix of the camera, P is the feature point set.

[0022] Preferably, the step of iteratively fitting the plane includes:

[0023] Randomly select three points using the RANSAC algorithm to establish an initial plane model, and screen the inliers within a distance threshold of 0.5;

[0024] Re-fit the plane parameters based on the least squares method, and complete the preliminary fitting after ten iterations;

[0025] Eliminate the outliers with excessive adjacent point distances through the PCA algorithm, and re-project them onto the fitted plane until the distance threshold of all points within the plane is less than 0.3 or the iteration diverges.

[0026] Preferably, adopt the iterative closest point algorithm optimized based on the mean square error, perform point cloud registration according to the optimized three-dimensional coordinate point set, convert the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three-way displacement change of the crack through the difference between the two measurement results, including:

[0027] Based on the generated optimized three-dimensional coordinate point set , calculate the centroids and of the three-dimensional points in the camera coordinate system and the target world coordinate system respectively;

[0028] According to the centroids and , perform centroid removal processing on each point to obtain the centroid-removed coordinates and ; Among them, and , and are respectively the actually observed two - dimensional image points;

[0029] Construct a covariance matrix , and perform singular value decomposition on to obtain the decomposition result; where, ; The formula for the decomposition result is: ; U and V are respectively the left singular vector and the right singular vector;

[0030] Calculate the optimal rotation matrix R’ and translation vector t’ according to the decomposition result; where, the calculation formula for the optimal rotation matrix R’ is: and translation vector ;

[0031] Use the optimal rotation matrix R’ and translation vector t’ to transform the set of three - dimensional coordinate points in the camera coordinate system obtained by solving to the world coordinate system, and obtain the transformed set of coordinate points ;

[0032] Calculate the root - mean - square error between the optimized set of three - dimensional coordinate points and the transformed set of coordinate points ; ; where, , is the distance between the i th point in the source point cloud and the corresponding point in the target point cloud, n is the total number of matching point pairs;

[0033] Iterate the optimal rotation matrix R’ and translation vector t’ until the RMSE is less than the threshold 0.1 or the iteration diverges, and finally output the crack displacement change amount in the world coordinate system.

[0034] Preferably, the calculation formula for the crack displacement change amount is:

[0035]

[0036] where, is the crack displacement change amount, , and are respectively the X axis, Y axis andZ The coordinate values of the axis 、 and are respectively the three - dimensional coordinates of the crack after change; 、 and are respectively the three - dimensional coordinates of the crack before change.

[0037] A three - dimensional measurement research system for cracks based on the fusion of multiple coordinate systems, comprising:

[0038] An image input unit, configured to take two target arrangement photos from different positions through the same camera to obtain initial images;

[0039] An image pre - processing unit, configured to perform binarization and filtering processing on the initial images to obtain pre - processed images

[0040] A density clustering unit, configured to extract the concentric - circle edge information in the pre - processed images and obtain an initial feature point set through a density - based spatial clustering algorithm;

[0041] A center - point correction unit, configured to perform distortion correction on the initial feature point set based on the cross - ratio invariance to eliminate the elliptical projection error caused by non - parallel shooting and obtain a corrected feature point set;

[0042] A pose solution unit, configured to initially calculate the rotation matrix and translation vector for the two shootings, and optimize the reprojection error to less than the threshold of 0.1 in combination with the Levenberg - Marquardt method, and output a projection matrix;

[0043] A coordinate solution unit, configured to solve the initial three - dimensional coordinate point set in the camera coordinate system based on the internal parameter matrix of camera calibration and the projection matrix by using the least - squares method;

[0044] A plane fitting unit, configured to pre - screen the three - dimensional point cloud data in the initial three - dimensional coordinate point set through the random sample consensus algorithm, eliminate outliers in combination with principal component analysis, iteratively fit the plane and project to generate an optimized three - dimensional coordinate point set;

[0045] A coordinate system conversion unit, configured to perform point cloud registration according to the optimized three - dimensional coordinate point set by using an iterative closest point algorithm optimized based on the mean square error, convert the three - dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three - dimensional displacement change of the crack through the difference between the two measurement results.

[0046] According to the specific embodiments provided by the present invention, the following technical effects of the present invention are disclosed:

[0047] The present invention provides a method and system for three-dimensional measurement of cracks based on fusion of multiple coordinate systems. The method comprises: taking two target arrangement photos from different positions by the same camera to obtain an initial image; binarizing and filtering the initial image to obtain a preprocessed image; extracting the concentric circle edge information in the preprocessed image, and obtaining an initial feature point set by a density-based spatial clustering algorithm; performing distortion correction on the initial feature point set based on cross ratio invariance to eliminate the elliptical projection error caused by non-parallel shooting, and obtaining a corrected feature point set; preliminarily calculating the rotation matrix and translation vector of the two shots, and combining the Levenberg-Marquardt method Optimize the reprojection error to below the threshold of 0.1, and output the projection matrix; based on the camera calibration internal parameter matrix and the projection matrix, use the least squares method to solve the initial three-dimensional coordinate point set in the camera coordinate system; pre-screen the three-dimensional point cloud data in the initial three-dimensional coordinate point set by the random sampling consensus algorithm, remove outliers in combination with principal component analysis, iteratively fit the plane and project, and generate an optimized three-dimensional coordinate point set; use the iterative nearest point algorithm based on mean square error optimization, perform point cloud registration according to the optimized three-dimensional coordinate point set, convert the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three-dimensional displacement change of the crack by the difference between the two measurement results. The present invention aims to establish a connection between the world coordinate system and the pixel coordinate system by specifying the world coordinate system, solve the relative position between the cameras, and perform three-dimensional reconstruction to avoid the problem of missing depth information in monocular vision. In the way of establishing an equivalent model of the three-dimensional displacement of the crack, the change of the coordinates of the sub-plate in the world coordinate system is observed to achieve the three-dimensional displacement measurement of the crack. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0050] Figure 2 An algorithm flow chart provided for an embodiment of the present invention;

[0051] Figure 3 A plane fitting result diagram provided by an embodiment of the present invention;

[0052] Figure 4 A target arrangement diagram provided for an embodiment of the present invention;

[0053] Figure 5 Iterative flow chart for solving posture problems provided by an embodiment of the present invention;

[0054] Figure 6 The triaxial displacement change value in the X-axis unidirectional movement test result provided by the embodiment of the present invention;

[0055] Figure 7 The triaxial error value in the X-axis unidirectional movement test result provided by the embodiment of the present invention;

[0056] Figure 8 The X-axis movement stability in the X-axis unidirectional movement test result provided by the embodiment of the present invention;

[0057] Figure 9 The triaxial displacement change value in the Y-axis unidirectional movement test result provided by the embodiment of the present invention;

[0058] Figure 10 The triaxial error value in the Y-axis unidirectional movement test result provided by the embodiment of the present invention;

[0059] Figure 11 The Y-axis movement stability in the Y-axis unidirectional movement test result provided by the embodiment of the present invention;

[0060] Figure 12 The triaxial displacement change value in the Z-axis unidirectional movement test result provided by the embodiment of the present invention;

[0061] Figure 13 The triaxial error value in the Z-axis unidirectional movement test result provided by the embodiment of the present invention;

[0062] Figure 14 The Z-axis movement stability in the Z-axis unidirectional movement test result provided by the embodiment of the present invention;

[0063] Figure 15 The triaxial displacement change value in the test result when moving the XYZ three axes simultaneously provided by the embodiment of the present invention;

[0064] Figure 16 The triaxial error value in the test result when moving the XYZ three axes simultaneously provided by the embodiment of the present invention. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] The purpose of the present invention is to provide a method and system for three-dimensional measurement of cracks based on the fusion of multiple coordinate systems. By specifying the world coordinate system, the purpose is to establish a connection between it and the pixel coordinate system, solve the relative position between cameras, and perform three-dimensional reconstruction to avoid the problem of missing depth information in monocular vision. By establishing an equivalent model of three-dimensional displacement of cracks, the changes of the coordinates of the sub-plate in the world coordinate system are observed to achieve three-dimensional displacement measurement of cracks.

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a three-dimensional crack measurement research method based on multi-coordinate system fusion, including:

[0069] Step 100: Take two target arrangement photos from different positions using the same camera to obtain an initial image;

[0070] Step 200: binarizing and filtering the initial image to obtain a preprocessed image;

[0071] Step 300: extracting the edge information of concentric circles in the preprocessed image, and obtaining an initial feature point set by a density-based spatial clustering algorithm;

[0072] Step 400: performing distortion correction on the initial feature point set based on the cross ratio invariance to eliminate the elliptical projection error caused by non-parallel shooting, and obtaining a corrected feature point set;

[0073] Step 500: Preliminarily calculate the rotation matrix and translation vector of the two shots, and optimize the reprojection error to below a threshold of 0.1 by combining the Levenberg-Marquardt method, and output the projection matrix;

[0074] Step 600: Based on the camera calibration intrinsic parameter matrix and projection matrix, the initial three-dimensional coordinate point set in the camera coordinate system is solved by the least square method;

[0075] Step 700: pre-screening the 3D point cloud data in the initial 3D coordinate point set by using a random sampling consensus algorithm, removing outliers in combination with principal component analysis, iteratively fitting the plane and projecting it, and generating an optimized 3D coordinate point set;

[0076] Step 800: Using an iterative closest point algorithm based on mean square error optimization, perform point cloud registration according to the optimized three-dimensional coordinate point set, transform the three-dimensional coordinates in the camera coordinate system into the world coordinate system, and calculate the three-dimensional displacement change of the crack by the difference between the two measurement results.

[0077] Specifically, as Figure 2 shown, the present invention provides a research method for three-way measurement of cracks based on the fusion of multiple coordinate systems, including the following steps:

[0078] 1) Image input: Take two photos of the target layout from different positions with the same camera;

[0079] 2) Image preprocessing: First, perform preprocessing such as binarization and image filtering on the image;

[0080] 3) Density clustering: Then, use the canny operator to extract the edge information of the concentric circles for density clustering, and use DBSCAN density clustering to obtain the centers of the concentric circles, forming an initial feature point set P';

[0081] 4) Center correction: Since the on-site shooting is often not parallel shooting, the circles on the object plane will appear as ellipses when projected onto the image plane, resulting in an error between the center and the actual projection point, that is, the eccentricity error. To eliminate the eccentricity error, in this embodiment, the cross-ratio invariance is used to correct the distortion of the center. After sorting, the feature point set can be formed ;

[0082] 5) Solve the relative position of the computer: First, use the EPnP algorithm to initially obtain the rotation matrix R and translation vector t required for coordinate system conversion, obtain pi, tl, Rr, and tr for the two shootings, and then use the Levenberg-Marquardt method to optimize the rotation matrix R and displacement t vector of the camera until the reprojection error value is less than the given threshold of 0.1, and then stop the iteration and output the projection matrix M.

[0083] 6) Coordinate solution: According to the obtained feature point set , and the internal parameter matrix K obtained by camera calibration and the projection matrix M obtained in step 5, use the least squares method to solve the three-dimensional coordinate point set in the camera coordinate system.

[0084] 7) Plane fitting: Since the pose derivation is based on the ideal situation, after three-dimensional reconstruction, it is necessary to optimize the obtained coordinate point set to eliminate the influence of interference items such as noise. Considering that the target plane is basically horizontal, the theoretically solved three-dimensional point set should also be located on the same plane. Therefore, use plane fitting to compensate for image distortion and project onto the fitting plane to form a new data point set , and then project through the ICP algorithm and iterative optimization based on the mean square error, and transform it into the world coordinate system. Finally, an optimized three-dimensional coordinate point set is obtained. The plane fitting result is as Figure 3 .

[0085] 8) Coordinate system transformation: The ICP algorithm based on mean square error iterative optimization is used for point cloud registration to solve the pose, and the transformation relationship of the coordinates in the camera coordinate system to the coordinates in the world coordinate system is obtained. And the coordinates in the world coordinate system are obtained. , According to the coordinates of the right target measured twice before and after, Subtracting them can obtain the three-dimensional displacement change of the crack.

[0086]

[0087] Among them, is the displacement change of the crack, , and are respectively the X axis, Y axis and Z axis coordinate values of the displacement change of the crack, , and are respectively the three-dimensional coordinates of the crack after the change; , and are respectively the three-dimensional coordinates of the crack before the change.

[0088] Furthermore, for the target layout photo in step 1), in this embodiment, a concentric circle target is independently designed, and the concentric circle center point set is used as the feature point set. On the one hand, the extraction accuracy of the feature points can be improved through the constraint conditions brought by the concentric circles, and on the other hand, the error caused by incorrect feature point matching can be eliminated through sorting. The target layout diagram is as Figure 4 .

[0089] Furthermore, when the camera is imaging in step 5), for any point P in space, the two views in the left and right camera coordinate systems can be associated with . Then the following relationship can be obtained:

[0090]

[0091] Among them: , respectively represent the positions of point in the left and right camera coordinate systems; and or ( and ) represent the rotation matrix and translation vector from the camera to the space point on the left camera (right camera); and are the relative rotation matrix and translation vector between the left and right camera coordinate systems; Substituting the , , , Substitute it to solve the pose relationship, thus forming a projection matrix .

[0092] Furthermore, the coordinate conversion relationship between point D( , , ) in the world coordinate system and point P(u, v) in the pixel coordinate system in step 6) is as follows:

[0093]

[0094] In the formula, K is the internal parameter matrix of the camera, and Zc is the coordinate of the Z-axis in the camera coordinate system.

[0095] Furthermore, in order to prevent falling into the local optimum situation during plane fitting in step 6), in this embodiment, before plane fitting, the RANSAC algorithm is used for pre-fitting to screen the point cloud data. First, a plane is established by randomly selecting three points in the point cloud data, and the distances from all points in the point cloud set to the plane model are calculated, and 0.5 is used as the threshold to judge the boundary. When the number of inlier points meets the requirements, all points on the plane (inliers) are used to recalculate the plane parameters based on the least squares principle, and iterate ten times to complete the preliminary fitting.

[0096] Furthermore, after RANSAC pre-fitting in step 7), referring to the true distance set by the target, the points with relatively large distances from adjacent points in the three-dimensional point cloud are removed by using the PCA algorithm, and then plane fitting is performed again. After the fitting is completed, the previously removed points are re-projected onto the fitting plane. Iterate until the distance from all points in the plane to the plane is less than the set threshold of 0.3 or the iteration situation diverges. After the fitting is completed, a new coordinate point set can be obtained .

[0097] Furthermore, the specific steps for solving the pose in step 8) are as follows:

[0098] 1) Calculate the centroids of the 3D points in two different coordinate systems respectively and ;

[0099] 2) Calculate the centroid-removed coordinates of each point , ;

[0100] 3) Define the matrix , and then perform SVD decomposition on the matrix ;

[0101] 4) When W is full rank, calculate to obtain , ;

[0102] 5) Calculate the root mean square error:

[0103]

[0104]

[0105] where is the distance between the i-th point in the source point cloud and the corresponding point in the target point cloud, and n is the total number of matching point pairs.

[0106] 6) Set the root mean square threshold to 0.1, and iterate the transformation matrix until the threshold is less than 0.1 or the iteration diverges and stops iterating. The iteration flowchart is as Figure 5 .

[0107] Furthermore, the crack three-way measurement research method based on multi-coordinate system fusion proposed by the present invention has the characteristics of high detection accuracy and wider applicability compared with the existing solutions.

[0108] As an optional implementation manner, in this embodiment, a camera is used to take multi-angle photos to simulate the shooting process of a binocular camera. After obtaining two pictures, they are input into a computer terminal to calculate the required test parameters, and finally the test results are obtained. The camera model is Nikon D3100, with 14.2 million pixels, effective pixels of 4608×3702, sensor size of 23.1×15.4mm, image processor EXPEED 2, and focal length of 18 - 55mm

[0109] To verify the accuracy and stability of the algorithm, in this embodiment, based on a fixed sliding table, a three-way displacement accuracy and stability detection device is designed. The device consists of a horizontal rail, a vertical rail, a vertical rail, and a dial indicator. The left and right test targets are respectively fixed on the mobile end and the fixed end, and dial indicators are arranged in the sliding horizontal direction. The targets are moved using the rails to simulate the three-way changes of the crack. The details of the test equipment are as Figure 6 shown.

[0110] During the test, the test equipment is photographed from multiple angles using a camera, and the target is placed at the center as much as possible during shooting. In this embodiment, the relative displacement on both sides of the crack is simulated by moving the target on the mobile end, and the reading of the dial indicator is used as the reference value of the displacement. During the test process, the displacement amount is controlled to be 1mm each time, and it is moved a total of ten times. After each movement, the image is collected for data processing, and the difference between the three-dimensional coordinates obtained and the coordinates at the start of shooting is the change situation of the crack. The test results are shown in Figures 6 to 16 . As shown in Figures 6 to 8It can be seen that when the X-axis moves unidirectionally, the error values between the measured displacement and the actual displacement are all controlled within ±0.4 mm. At the same time, the X-direction error range is within ±0.4 mm, indicating that the accuracy and stability of the algorithm are reliable, and the error volatility does not increase with the increase of the reference displacement. From Figures 9 to 11 It can be seen that when the Y-axis moves unidirectionally, the error values between the measured displacement and the actual displacement are all controlled within ±0.4 mm. At the same time, the Y-direction error range is within ±0.4 mm, indicating that the accuracy and stability of the algorithm are reliable, and the error volatility does not increase with the increase of the reference displacement. From Figures 12 to 14 It can be seen that when the Z-axis moves unidirectionally, the error values between the measured displacement and the actual displacement are all controlled within ±0.4 mm. At the same time, the Z-direction error range is within ±0.4 mm, indicating that the accuracy and stability of the algorithm are reliable, and the error volatility does not increase with the increase of the reference displacement. Since cracks will not only have unidirectional displacement in actual situations, the three-axis movement test is also necessary. Figure 15 and Figure 16 are the test results under the condition of simultaneously moving the X, Y, and Z axes. From Figures 15 to 16 It can be seen that when the three axes move, the error values between the measured displacement and the actual displacement are all controlled within ±0.4 mm. At the same time, the three-direction error range is within ±0.4 mm, indicating that the accuracy and stability of the algorithm are reliable, and the error change ranges of each axis are all within the same level, indicating that the measurement effect of the algorithm is good.

[0111] Analyzing the above test results, it can be seen that in the single-axis movement test, the displacement distance of the moving axis measured by the algorithm shows a linear change trend with the actual moving distance, and the absolute error of the fixed axis remains at a level similar to that of the measured axis, fluctuating within ±0.4 mm, indicating that the algorithm can accurately judge the change situation of the moving axis. In the three-axis movement test, the error still remains within ±0.4 mm.

[0112] The algorithm of this embodiment is analyzed as follows:

[0113] Table 1 Algorithm Resolution Analysis

[0114]

[0115] It can be seen from Table 1 that the fluctuation ranges of the three-axis measurements of the algorithm are all within 0.6 mm, proving that the resolution of the algorithm is within the controllable range.

[0116] Table 2 Algorithm Accuracy Analysis

[0117]

[0118] It can be seen from Table 2 that during the test, the absolute errors of the three axes are all within ±0.4 mm, meeting the requirements for crack measurement in the "Technical Specification for Safety Monitoring of Earth-Rock Dams" (±0.5 mm).

[0119] Table 3 Algorithm Repeatability Analysis

[0120]

[0121] As can be seen from Table 3, the maximum deviation of the algorithm is 0.4 mm on the x-axis, which proves that the repeatability of this embodiment is reliable.

[0122] Corresponding to the above method, this embodiment also provides a crack three-way measurement research system based on multi-coordinate system fusion, including:

[0123] An image input unit, configured to take two target arrangement photos from different positions through the same camera to obtain initial images;

[0124] An image preprocessing unit, configured to perform binarization and filtering processing on the initial images to obtain preprocessed images

[0125] A density clustering unit, configured to extract concentric circle edge information in the preprocessed images, and obtain an initial feature point set through a density-based spatial clustering algorithm;

[0126] A center correction unit, configured to perform distortion correction on the initial feature point set based on cross-ratio invariance to eliminate the elliptical projection error caused by non-parallel shooting, and obtain a corrected feature point set;

[0127] A pose solving unit, configured to initially calculate the rotation matrix and translation vector of the two shootings, and optimize the reprojection error to below the threshold of 0.1 in combination with the Levenberg-Marquardt method, and output a projection matrix;

[0128] A coordinate solving unit, configured to solve the initial three-dimensional coordinate point set in the camera coordinate system based on the internal parameter matrix of camera calibration and the projection matrix by using the least squares method;

[0129] A plane fitting unit, configured to pre-screen the three-dimensional point cloud data in the initial three-dimensional coordinate point set through the random sample consensus algorithm, eliminate outliers in combination with principal component analysis, iteratively fit the plane and project, and generate an optimized three-dimensional coordinate point set;

[0130] A coordinate system conversion unit, configured to perform point cloud registration according to the optimized three-dimensional coordinate point set by using the iterative closest point algorithm optimized based on the mean square error, convert the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three-way displacement change of the crack through the difference between the two measurement results.

[0131] The beneficial effects of the present invention are as follows:

[0132] (1) First, by specifying the world coordinate system, the EPnP algorithm establishes the connection between the world coordinate system and the pixel coordinate system, solves the relative position between the cameras, and performs three-dimensional reconstruction, thereby avoiding the problem of missing depth information in monocular vision. This solves the problem that traditional crack detection methods have high requirements for equipment and shooting posture and cannot be well adapted to on-site crack measurement.

[0133] (2) Secondly, the present invention independently designs a concentric circle target and uses the concentric circle center point set as the feature point set. On the one hand, the feature point extraction accuracy can be improved by the constraints of the concentric circle itself, and on the other hand, the error caused by the feature point matching error can be eliminated by sorting.

[0134] (3) The present invention designs a concentric dot matrix target and establishes a crack equivalent displacement model, which converts the three-dimensional change measurement problem of the crack into the relative change measurement problem of the target main plate and the sub-plate. The feature point set is obtained by camera shooting, and the sub-plate point set is transferred to the world coordinate system through three-dimensional reconstruction and coordinate system conversion. The three-dimensional change of the crack can be obtained by observing the change of the sub-plate coordinates. This solves the problem of lack of depth information in monocular vision measurement, and the problem that it cannot be used to quickly and accurately obtain the three-dimensional displacement of the crack.

[0135] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0136] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A research method for three-way measurement of cracks based on the fusion of multiple coordinate systems, characterized in that Including: Taking two photos of the target layout from different positions by the same camera to obtain initial images; Performing binarization and filtering on the initial images to obtain preprocessed images; Extracting concentric circle edge information in the preprocessed images and obtaining an initial feature point set through a density-based spatial clustering algorithm; Performing distortion correction on the initial feature point set based on cross-ratio invariance to eliminate the elliptical projection error caused by non-parallel shooting, and obtaining a corrected feature point set; Preliminarily calculating the rotation matrix and translation vector for the two shootings, and combining with the Levenberg-Marquardt method to optimize the reprojection error to below the threshold of 0.1, and outputting the projection matrix; Based on the internal parameter matrix of camera calibration and the projection matrix, using the least squares method to solve the initial three-dimensional coordinate point set in the camera coordinate system; Pre-screening the three-dimensional point cloud data in the initial three-dimensional coordinate point set through the Random Sample Consensus algorithm, combining with principal component analysis to remove outliers, iteratively fitting a plane and projecting to generate an optimized three-dimensional coordinate point set; Adopting an Iterative Closest Point algorithm optimized based on mean square error, performing point cloud registration according to the optimized three-dimensional coordinate point set, converting the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculating the three-way displacement change of the crack through the difference between the two measurement results; Preliminarily calculating the rotation matrix and translation vector for the two shootings, and combining with the Levenberg-Marquardt method to optimize the reprojection error to below the threshold of 0.1, and outputting the projection matrix, including: Establish the mapping relationship between the world coordinate system and the pixel coordinate system through the EPnP algorithm, and solve the rotation matrix in the left and right camera coordinate systems R l 、 R r and the translation vector t l 、 t r ; Substitute R l , R r , t l , t r into the formula to relate the left and right camera coordinate systems, obtaining the relative rotation matrix and translation vector R and translation vector t; Iteratively optimizing R and t through the Levenberg-Marquardt method until the reprojection error is less than 0.1, and outputting the projection matrix M; The expression of the mapping relationship between the world coordinate system and the pixel coordinate system is: Among them, , and are the coordinate values of the , , -axes, X -axis, Y -axis and Z -axis of point D( u and v are the coordinate values of the u , v )'s X -axis and Y -axis in the pixel coordinate system, Z c is the coordinate value of the Z -axis in the camera coordinate system, K is the internal parameter matrix of the camera, P is the set of feature points;​​ Adopting an Iterative Closest Point algorithm optimized based on mean square error, performing point cloud registration according to the optimized three-dimensional coordinate point set, converting the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculating the three-way displacement change of the crack through the difference between the two measurement results, including: Based on the generated optimized three-dimensional coordinate point set , calculate the centroids of the three-dimensional points in the camera coordinate system and the target world coordinate system respectively and ; According to the centroid and , each point is de-centered to obtain the de-centered coordinates and ; where and , and are respectively the actually observed two-dimensional image points and the obtained two-dimensional image points Construct a covariance matrix , and perform singular value decomposition on to obtain the decomposition result; where ; The formula for the decomposition result is: ; U and V are the left singular vector and the right singular vector respectively; Calculate the optimal rotation matrix according to the decomposition result R’ and translation vector t’ ; among them, the calculation formula for the optimal rotation matrix R' is: and translation vector ; Using the optimal rotation matrix R’ and translation vector t’ , the three-dimensional coordinate point set in the camera coordinate system obtained by solving is converted to the world coordinate system to obtain the converted coordinate point set ; The optimized three-dimensional coordinate point set after calculation and the transformed coordinate point set The root mean square error between ; The formula is as follows: ; Among them, , is the distance between the i th point in the source point cloud and the corresponding point in the target point cloud, n is the total number of matching point pairs; Iteratively optimal rotation matrix R’ and translation vector t’ , until the RMSE is less than the threshold of 0.1 or the iteration diverges, and finally output the crack displacement change in the world coordinate system; The calculation formula for the crack displacement change is: Among them, is the crack displacement change amount, , and are the coordinate values of the X axis, Y axis and Z axis of the crack displacement change amount respectively, , and are the three-dimensional coordinates of the crack after change respectively; , and are the three-dimensional coordinates of the crack before change respectively.

2. The research method for three - dimensional crack measurement based on multi - coordinate system fusion according to claim 1, characterized in that, The target is a preset concentric circle target, and the center point set of the concentric circles is used as the feature point set.

3. The method for crack three-dimensional measurement research based on multi-coordinate system fusion according to claim 1, characterized in that The steps of the iterative plane fitting include: Randomly selecting three points by the RANSAC algorithm to establish an initial plane model, and screening inliers within a distance threshold of 0.5; Re-fitting the plane parameters based on the least squares method, and iterating ten times to complete the preliminary fitting; Removing outliers with excessive adjacent point distances through the PCA algorithm, and re-projecting to the fitted plane until the distance threshold of all points in the plane is less than 0.3 or the iteration diverges.

4. A crack three-dimensional measurement research system based on the fusion of multiple coordinate systems, characterized in that, For implementing the method described in any one of claims 1 to 3, the system includes: An image input unit for taking two photos of the target layout from different positions by the same camera to obtain initial images; An image preprocessing unit for performing binarization and filtering on the initial images to obtain preprocessed images A density clustering unit for extracting concentric circle edge information in the preprocessed images and obtaining an initial feature point set through a density-based spatial clustering algorithm; A circle center correction unit, used for performing distortion correction on the initial feature point set based on cross ratio invariance to eliminate the ellipse projection error caused by non-parallel shooting, and obtaining a corrected feature point set; The pose solving unit is used to preliminarily calculate the rotation matrix and translation vector of the two shots, and optimize the reprojection error to below the threshold of 0.1 by combining the Levenberg-Marquardt method, and output the projection matrix; A coordinate solving unit, used to solve the initial three-dimensional coordinate point set in the camera coordinate system by using the least square method based on the camera calibration intrinsic parameter matrix and the projection matrix; A plane fitting unit is used to pre-screen the three-dimensional point cloud data in the initial three-dimensional coordinate point set by using a random sampling consensus algorithm, remove outliers in combination with principal component analysis, iteratively fit the plane and project it, and generate an optimized three-dimensional coordinate point set; The coordinate system conversion unit is used to adopt an iterative closest point algorithm based on mean square error optimization, perform point cloud registration according to the optimized three-dimensional coordinate point set, convert the three-dimensional coordinates in the camera coordinate system to the world coordinate system, and calculate the three-dimensional displacement change of the crack by the difference between two measurement results.

Images