A segment corner reconstruction method and system based on scanning data

Abstract

The present application relates to the field of bridge engineering, and particularly relates to a segment corner reconstruction method and system based on scanning data, the method comprising: based on the three-dimensional point cloud of the bridge segment and the surface patch model corresponding to the point cloud, respectively performing edge line fitting and reconstruction and surface fitting and reconstruction; determining the bridge segment corner point according to the corner point obtained by edge line fitting reconstruction and the corner point obtained by surface fitting reconstruction, the present application improves the success rate of corner feature extraction by the reconstruction fitting method under the condition of missing part of the scanning point cloud in the prior art, reduces the work difficulty of the early scanning operation, improves the operation success rate, avoids multiple scanning, and provides different reconstruction fitting methods according to the common corner error sources in the field, so as to adapt to the corner error extraction under different interference conditions.

Description

Technical Field

[0001] This invention relates to the field of bridge engineering, and in particular to a method and system for reconstructing segment corners based on scan data. Background Technology

[0002] Large and complex bridges require extremely high precision in matching and aligning the dimensions of multiple segments during actual assembly and hoisting. Traditional methods typically rely on manual on-site measurements, which are not only time-consuming and labor-intensive, but also prone to large measurement errors. Furthermore, interference from additional components can necessitate indirect measurements, further increasing errors and impacting construction progress and quality.

[0003] Existing technologies mostly employ laser-based or oblique photogrammetry scanning methods, which can relatively reduce absolute measurement deviations and avoid component interference. However, the processing of scanned point cloud data is a critical issue; due to interference or occlusion, some scanned point clouds are often missing. Among these, the measurement of corner points at component corners is a key measurement point data under existing construction methods and specifications. However, because segmental component corners often have features affecting the corners such as bevels, welds, and mating parts, existing corner point selection methods suffer from measurement difficulties, poor accuracy, and large errors. Consequently, manual confirmation is required during the reconstruction of segmental corners. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems of low success rate of scanning operations and insufficient corner measurement in the prior art, and to provide a segment corner reconstruction method and system based on scanning data.

[0005] In a first aspect, the present invention provides a segment corner reconstruction method based on scan data, comprising:

[0006] Based on the 3D point cloud of bridge segments and the corresponding surface model of the point cloud, edge fitting and reconstruction and surface fitting and reconstruction are performed respectively.

[0007] The corner points of the bridge segments are determined based on the corner points obtained by edge fitting and reconstruction and the corner points obtained by surface fitting and reconstruction.

[0008] The distance between the corner points of the bridge segment obtained by edge fitting reconstruction and the corner points obtained by surface fitting reconstruction is determined.

[0009] For common sources of corner errors (edges / surfaces) in the field, different reconstruction and fitting methods are provided for different types to adapt to corner error extraction under different interference conditions. The reconstruction and fitting methods improve the success rate of corner feature extraction, reduce the difficulty of the initial scanning work, improve the success rate of the operation, and avoid multiple scans.

[0010] Preferably, the contour of the patch model is used to downsample the point cloud after coarse positioning transformation to obtain multiple straight line edge groups;

[0011] Point-line matching is performed based on the distance from the midpoint of the point cloud after coarse localization transformation to the lines in the multi-line edge group;

[0012] The matched points are fitted to the line and projected onto a plane perpendicular to the corresponding line. The projected points are then clustered to obtain a cluster of line points.

[0013] Preferably, the edge reconstruction process includes:

[0014] The NURBS curves are used to reconstruct the cluster of straight line points to obtain the reconstructed curve group;

[0015] Extend both ends of the reconstructed curve group to determine the closest point between two adjacent extended curves;

[0016] The average value of the nearest points is taken to obtain the corner points under edge fitting and reconstruction.

[0017] A complete edge fitting and reconstruction scheme is provided to adapt to corner error extraction under edge interference conditions.

[0018] Preferably, the surface fitting process includes:

[0019] The point cloud after coarse localization transformation is filtered in a spherical space centered at the corner of the contour to obtain the disordered point cloud at the corner.

[0020] Determine the features of point p in the unordered point cloud at the corner, perform breadth-first clustering, and output the clustering container;

[0021] Each clustering container corresponds to a planar point cloud.

[0022] Preferably, the process of determining the features of point p in the disordered point cloud at the corner includes:

[0023] Determine the distance between a point p in the unordered point cloud at the corner and its nearest neighbor points, and take the points whose distance is less than the radius threshold as the threshold point set;

[0024] Calculate the covariance matrix of the threshold point set, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0025] Based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, the flattening index and normal of point p are determined as features of point p.

[0026] Preferably, the process of performing breadth-first clustering on point p includes:

[0027] Sort all points p according to the flatness index;

[0028] The sorted point sets are placed into an empty clustering container and numbered.

[0029] Based on the normal of point p and all its neighboring points, cluster each point p corresponding to a given number. After clustering is complete, output the clustering container.

[0030] Preferably, for any two planar point clouds Pa and Pb, the surface reconstruction process includes:

[0031] S101. Based on the planar distance threshold, obtain the optimal planes ha and hb corresponding to the planar point clouds Pa and Pb according to the RANSAC algorithm;

[0032] S102. Determine the intersection line lab of the optimal planes ha and hb;

[0033] S103. Project the planar point clouds Pa and Pb using the columnar filtering range, calculate the distance from the projected points to the intersection line lab, and retain the points projected within the columnar filtering range.

[0034] S104. Determine the optimal plane corresponding to the retained points, and calculate the intersection line between the optimal planes;

[0035] S105. Change the columnar filtering range and repeat S103 and S104 to obtain several sets of projection intersection results.

[0036] S106. Select the optimal intersection line from several sets of projection intersection line results, and obtain the corner points under surface fitting and reconstruction based on the optimal intersection line.

[0037] Preferably, the method for determining corner points under surface fitting and reconstruction is as follows:

[0038] After filtering out the optimal intersection lines using the normal distribution and applying a weighted average, k-clustering is performed.

[0039] The corner points under surface fitting and reconstruction are determined based on the distance between the cluster center and the surrounding original point cloud cluster centers;

[0040] The optimal intersection line is determined by clustering several sets of projected intersection line results; or...

[0041] The ratio of the interior points contained in the two planes corresponding to the projection intersection line is determined.

[0042] Preferably, the method for determining the corner points of bridge segments is as follows:

[0043] Perform multiple surface reconstructions to obtain possible edge and corner point groups during the surface reconstruction process;

[0044] Cluster the corner points to obtain the cluster mean points;

[0045] Compare the distances between the cluster mean and the corner points under edge fitting and reconstruction:

[0046] If the value is less than the error, the cluster mean point and the corner points under the edge fitting and reconstruction are taken as the corner points of the bridge segment.

[0047] If the error is greater than or equal to the error, the result is determined manually based on all points in the corner point group, the cluster mean point, the corner points under edge fitting and reconstruction, and all points within the preset perimeter of the corner points under edge fitting and reconstruction.

[0048] In a second aspect, the present invention provides a segmental corner reconstruction system based on scan data for performing the method described in the first aspect, comprising:

[0049] The data import module is used to read the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud;

[0050] The edge fitting and reconstruction module is used to perform edge fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under edge fitting and reconstruction.

[0051] The surface fitting and reconstruction module is used to perform surface fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under surface fitting and reconstruction.

[0052] The bridge segment corner point determination module is used to determine the corner points of bridge segments based on the corner points under edge fitting and reconstruction and the corner points under surface fitting and reconstruction.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0054] 1. This invention provides different reconstruction and fitting methods for corner points and surfaces obtained through edge fitting and reconstruction, targeting common corner error sources in the field, in order to adapt to corner error extraction under different interference conditions;

[0055] 2. This invention improves the success rate of corner feature extraction by using a reconstruction fitting method. In the case of missing part of the scanned point cloud (due to interference or occlusion, etc.) in the existing technology, it reduces the difficulty of the initial scanning operation, improves the success rate of the operation, and avoids multiple scans. Attached Figure Description

[0056] Figure 1 Here is a flowchart of the segment corner reconstruction method based on scan data;

[0057] Figure 2 A schematic diagram of edge fitting and reconstruction;

[0058] Figure 3 This is a schematic diagram after surface fitting. Detailed Implementation

[0059] The present invention will now be described in further detail with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0060] Example 1

[0061] Please refer to Figure 1 As shown, this embodiment provides a segment corner reconstruction method based on scan data, including:

[0062] Based on the 3D point cloud of bridge segments and the corresponding surface model of the point cloud, edge fitting and reconstruction and surface fitting and reconstruction are performed respectively.

[0063] The corner points of the bridge segments are determined based on the corner points obtained by edge fitting and reconstruction, as well as the corner points obtained by surface fitting and reconstruction.

[0064] Example 2

[0065] Please refer to Figure 1 , Figure 2 as well as Figure 3 This embodiment is one or more preferred solutions of Embodiment 1, and is further optimized based on Embodiment 1.

[0066] Preferably, to increase the accuracy of the scheme, before performing edge fitting and reconstruction and surface fitting and reconstruction, a coarse positioning transformation is performed on the three-dimensional point cloud of the bridge segment and the surface model corresponding to the point cloud.

[0067] Performing a coarse positioning transformation before formal fitting and reconstruction can effectively reduce the accuracy pressure in subsequent steps.

[0068] Specifically, the above scheme can be expressed as: performing a coarse localization transformation on the scanned point cloud P0 and the surface M0 to obtain the rotation transformation matrix M between the point cloud P0 and the surface M0. R Following this rotation transformation, the coarsely positioned point cloud P1 is obtained.

[0069] As a preferred embodiment, after the rotation transformation, the nearest point P between each point in the scanned point cloud P0 and the triangular facet M0 is calculated. i To obtain P0 and P i ICP transformation matrix M Ri After applying this transformation, repeat the above steps until P0 and P... i The average distance no longer decreases, and the final transformation matrix M is obtained. result .

[0070] As a preferred embodiment, the calculation of the nearest point between the point cloud and the face can be accelerated by constructing an octree of the M0 triangular face.

[0071] As a preferred embodiment, SVD decomposition can also be used in the above-mentioned calculation of the scanned point cloud and the nearest point to accelerate the calculation.

[0072] The edge fitting and reconstruction process can be as follows: Figure 2 As shown, the point cloud was registered with the original model. The intersection of the segmentation surfaces near the corner points of the original model was used as the recommended corner points, and the lines connecting the recommended corner points at each corner position were the fitted edges. Specifically:

[0073] The edge fitting process includes:

[0074] The point cloud after coarse localization transformation is downsampled using the upper and lower contours of the patch model to obtain multiple straight line edge groups;

[0075] As a preferred embodiment, when downsampling the point cloud P1 after coarse positioning using the upper and lower contours PL0, since there are many interference factors at the corners of the scanned point cloud, the corners within the d0 range of the upper and lower contours PL0 (generally 50-200mm) can be deleted before this step to obtain the multiple straight edge group PL1 with interference factors removed.

[0076] Point-line matching is performed based on the distance from the midpoint of the point cloud after coarse localization transformation to the lines in the multi-line edge group;

[0077] As a preferred embodiment, the scanned point cloud can be filtered before point-line fitting; firstly, the multiple straight lines on each side are divided into several small line segments according to the distance d2 (generally 2-5mm), and a cube with length, width and height of d2 is established with the direction of the line segment as the main direction. The point cloud of P1 is grouped according to the cube, thereby constructing a cvMat that can be computed in parallel.

[0078] The matched points are fitted to the line and projected onto a plane perpendicular to the corresponding line. The projected points are then clustered to obtain a cluster of line points.

[0079] As a preferred embodiment, if a parallel computing-enabled cvMat has been constructed previously, all points within each cube are projected onto the face of the cube perpendicular to the direction of the line segment, and the projected points are clustered to obtain a cluster of line points.

[0080] The edge reconstruction process includes:

[0081] The straight line point clusters are reconstructed using NURBS (Non-Uniform Rational B-Splines) curves to obtain the reconstructed curve set;

[0082] The points in the straight line point cluster are the interpolation points of this NURBS curve (i.e., the reconstructed curve passes through these points). Generally, the order of the curve is set to the number of interpolation points, but it can also be simplified to any value above 5 according to actual needs. The above operation is performed on each edge to obtain the initial reconstructed curve group C0.

[0083] Extend both ends of the reconstructed curve group to determine the closest point between two adjacent extended curves;

[0084] The average value of the nearest points is taken to obtain the corner points under edge fitting and reconstruction.

[0085] Since a NURBS curve is essentially a mathematical equation, according to its definition, it is extended at both ends. The closest points on any two adjacent extensions are found, and the average of these two points is taken to obtain a point. This point serves as the starting point of the two adjacent curves and also as the corner point for edge fitting and reconstruction, denoted as P. cornero .

[0086] The surface fitting process includes:

[0087] The point cloud after coarse positioning transformation is filtered in a spherical space centered at the corner of the ideal contour to obtain the disordered point cloud at the corner.

[0088] As one possible implementation, the filtering method is as follows: filter according to the threshold d of the spherical space, and filter out point clouds whose points are more than d relative to the center of the sphere;

[0089] Determine point p in the disordered point cloud at the corner (here, point p is the same as P). corner The characteristics of the clustering are analyzed, and breadth-first clustering is performed to output the clustering container.

[0090] Each clustering container corresponds to a planar point cloud.

[0091] The result after surface fitting can be as follows Figure 3 As shown, all points are the original point cloud filtered by spatial radius and segmented according to surface clustering. Different colors represent different segmentation surfaces. The smaller the threshold of the spherical space, the more surfaces are segmented.

[0092] As a preferred embodiment, the process of determining the features of point p in the disordered point cloud at the corner includes:

[0093] Determine the distance between a point p in the unordered point cloud at the corner and its nearest neighbor points, and take the points whose distance is less than the radius threshold as the threshold point set;

[0094] Calculate the covariance matrix of the threshold point set, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0095] Based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, the flattening index and normal of point p are determined as features of point p.

[0096] Among them, the flatness index α is the smallest eigenvalue divided by the sum of all eigenvalues; the normal is the eigenvector corresponding to the smallest eigenvalue; in addition, the set of indices of all neighbors of point p can be recorded to prepare for the subsequent breadth-first clustering process.

[0097] The process of performing breadth-first clustering on point p includes:

[0098] Sort all points p according to the flatness index;

[0099] The sorted point sets are placed into an empty clustering container and numbered.

[0100] Based on the normal of point p and all its neighboring points, cluster each point p corresponding to a number. After the clustering is completed, output the clustering container.

[0101] Each clustering container corresponds to a planar point cloud.

[0102] As a preferred embodiment, the process of performing breadth-first clustering on point p may specifically include:

[0103] a. Sort all three-dimensional points p according to their flatness index α in ascending order to form the point sequence P. plane ;

[0104] b. Maintain a usage record list for all points p, initialized to all unused; and create an empty clustering container C, which holds the sorted point set P. plane Add the first unused point in container C to container C, and set the seed point to the last point added to container C.

[0105] c. Start clustering from the currently set seed point and record the index j of the seed point in the current container;

[0106] Step c specifically also includes:

[0107] i. Denote the seed point and all its neighboring points as the local point set P. part ;

[0108] ii. Calculate P part Record the distance d_point from all points to the seed point, and record the median distance d_point_median;

[0109] iii. Calculate P part p_mean is the average point (center point) of all points;

[0110] iv. Convert the center point p_mean and the normal n_p of the seed point into a plane denoted as h;

[0111] v. Calculate P part Record the distance d_plane (i.e., unsigned projected distance) from all points to plane h, and record the median distance d_plane_median;

[0112] vi. Calculate the absolute value of the difference between the planar distance d_plane and the median d_plane_median for each point, and record the median of these values, diff_d_plane_median;

[0113] vii. Set an upper limit for the distance from a point to a plane, d_plane_max, to filter out points whose geometric distance to the fitted plane exceeds the upper limit for the plane distance, d_plane_max.

[0114] d_plane_max = d_plane_median + 2 diff_d_plane_medianv b

[0115] Where b = 1.4826 is the absolute median difference constant;

[0116] viii. Iterate through all neighboring points of the seed point, where for each neighboring point q:

[0117] If q has already been used, proceed to step ix;

[0118] If the planar distance d_plane corresponding to q is greater than or equal to the upper limit of the planar distance threshold d_plane_max, then proceed to step ix;

[0119] If the point distance d_point corresponding to q is greater than or equal to the median point distance d_point_median, then proceed to step ix;

[0120] Calculate the angle θ between the normal n_q of point q and the normal n_p of the seed point;

[0121] ix. Given a merging angle threshold:

[0122] If both 180-θ and θ are greater than the angle threshold, proceed to step d;

[0123] Add point q to the clustering container C and mark point q as used. If θ > 90°, mirror the normal of point q, that is, multiply the normal of point q by one.

[0124] d. Check if index j+1 exists in container C:

[0125] If it does not exist, check the number of points in container C. If the number is greater than the minimum allowed number of points in the point set, size_C_max, then submit the clustering and return to step b.

[0126] If it exists, set the seed point to the point with index j+1 in the container and return to step c.

[0127] For any two planar point clouds Pa and Pb, the surface reconstruction process includes:

[0128] S101. Based on the planar distance threshold, obtain the optimal planes ha and hb corresponding to the planar point clouds Pa and Pb according to the RANSAC algorithm;

[0129] S102. Determine the intersection line lab of the optimal planes ha and hb;

[0130] The parameterized equation corresponding to the intersection line lab is lab = vab + t × sab;

[0131] Where, vab is a point on the intersection line, sab is the direction of the intersection line, t is any real number, the intersection line lab belongs to both plane ha and hb, and the direction of the intersection line is perpendicular to both the plane normals nha and nhb.

[0132] S103. Project the planar point clouds Pa and Pb using the columnar filtering range, calculate the distance from the projected points to the intersection line lab, and retain the points projected within the columnar filtering range.

[0133] The range of columnar filtering can be represented as [r_min, r_max];

[0134] S104. Determine the optimal plane corresponding to the retained points, and calculate the intersection line between the optimal planes;

[0135] S105. Change the columnar filtering range and repeat S103 and S104 to obtain several sets of projection intersection results.

[0136] S106. Select the optimal intersection line from several sets of projection intersection line results, and obtain the corner point P under surface fitting and reconstruction based on the optimal intersection line. face .

[0137] The optimal intersection line is determined by clustering several sets of projected intersection line results; or...

[0138] The ratio of the interior points contained in the two planes corresponding to the projection intersection line is determined.

[0139] As one possible approach, the optimal intersection line is the intersection line of the two planes whose interior points contain the highest proportion of the two planes corresponding to the projection intersection line.

[0140] As another possible implementation, the optimal intersection line can be determined according to other rules or by manual judgment.

[0141] As one possible implementation, several sets of projection intersection results (i.e., optimal intersection results) are filtered using a normal distribution and then weighted and averaged. Based on this, k-clustering is performed, and the distance between the cluster centers and the original point cloud cluster centers on the surrounding original point clouds is calculated to obtain the corner point P under surface fitting and reconstruction. face .

[0142] As one possible implementation, when determining the corner points of bridge segments, the corner points P under surface fitting and reconstruction are used. face Corner point P under edge fitting and reconstruction cornero The distance is determined if P face With P cornero If the distance is less than the allowable error ξ, then P is used directly. face With P cornero The mean point is used as the corner point.

[0143] As a preferred embodiment, when determining the corner points of bridge segments, due to the existence of errors, different combinations of values ​​(such as different planar thresholds and columnar filtering ranges, different thresholds d of spherical space and different radii R of spherical space, etc.) can be tried to perform surface reconstruction, so as to obtain possible corner point groups P during the surface reconstruction process. selcet Cluster these corner points and obtain the mean and distribution of the cluster with the largest number of points. If the mean point P of this cluster is... face With P cornero If the distance is less than the allowable error ξ (e.g., 2mm), then P is used directly. face With P cornero The mean point is used as the corner point. If this distance is greater than the allowable error ξ, then all P values ​​are output. selcet、 P face、 P cornero All points within a d5 range around the device, typically 200mm, are displayed at the front end, and a suitable corner point is selected manually from this range.

[0144] Example 3

[0145] This embodiment provides a segment corner reconstruction system based on scan data, used to execute the segment corner reconstruction system method based on scan data described in either Embodiment 1 or Embodiment 2, including:

[0146] The data import module is used to read the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud;

[0147] The edge fitting and reconstruction module is used to perform edge fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under edge fitting and reconstruction.

[0148] The surface fitting and reconstruction module is used to perform surface fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under surface fitting and reconstruction.

[0149] The bridge segment corner point determination module is used to determine the corner points of bridge segments based on the corner points under edge fitting and reconstruction and the corner points under surface fitting and reconstruction.

[0150] Among them, the bridge segment corner point determination module, due to the existence of errors when determining the corner points of bridge segments, can also try different combinations of values ​​to perform surface reconstruction, so as to obtain the possible corner point group P during the surface reconstruction process. selcet Cluster these corner points and obtain the mean and distribution of the cluster with the largest number of points. If the mean point P of this cluster is... face With P cornero If the distance is less than the allowable error ξ (e.g., 2mm), then P is used directly. face With P cornero The mean point is used as the corner point. If this distance is greater than the allowable error ξ, then all P values ​​are output. selcet、 P face、 P cornero and its P cornero All points within a preset range d5, typically 200mm, are displayed at the front end, and a suitable corner point is selected manually from this range.

[0151] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A segment corner reconstruction method based on scan data, characterized in that, include: Based on the 3D point cloud of bridge segments and the corresponding surface model of the point cloud, edge fitting and reconstruction and surface fitting and reconstruction are performed respectively. The corner points of the bridge segment are determined based on the corner points obtained by edge fitting and reconstruction and the corner points obtained by surface fitting and reconstruction. The corner points of the bridge segment are determined by the distance between the corner points obtained by edge fitting and reconstruction and the corner points obtained by surface fitting and reconstruction. The method for determining the corner points of bridge segments is as follows: Perform multiple surface reconstructions to obtain possible edge and corner point groups during the surface reconstruction process; Cluster the corner points to obtain the cluster mean points; Compare the distances between the cluster mean and the corner points under edge fitting and reconstruction: If the value is less than the error, the cluster mean point and the corner points under the edge fitting and reconstruction are taken as the corner points of the bridge segment. If the error is greater than or equal to the error, the result is determined manually based on all points in the corner point group, the cluster mean point, the corner points under edge fitting and reconstruction, and all points within the preset perimeter of the corner points under edge fitting and reconstruction.

2. The segment corner reconstruction method based on scan data according to claim 1, characterized in that, The edge fitting process includes: The contours of the patch model are used to downsample the point cloud after coarse localization transformation to obtain multiple straight line edge groups; Point-line matching is performed based on the distance from the midpoint of the point cloud after coarse localization transformation to the lines in the multi-line edge group; The matched points are fitted to the line and projected onto a plane perpendicular to the corresponding line. The projected points are then clustered to obtain a cluster of line points.

3. The segment corner reconstruction method based on scan data according to claim 2, characterized in that, The edge reconstruction process includes: The NURBS curves are used to reconstruct the cluster of straight line points to obtain the reconstructed curve group; Extend both ends of the reconstructed curve group to determine the closest point between two adjacent extended curves; The average value of the nearest points is taken to obtain the corner points under edge fitting and reconstruction.

4. The segment corner reconstruction method based on scan data according to claim 1, characterized in that, The surface fitting process includes: Within a spherical space centered at the contour corner, the point cloud after coarse localization transformation is filtered to obtain an unordered point cloud at the corner; the features of point p in the unordered point cloud at the corner are determined, and breadth-first clustering is performed to output a clustering container; where each clustering container corresponds to a planar point cloud.

5. The segment corner reconstruction method based on scan data according to claim 4, characterized in that, The process of determining the features of point p in an unordered point cloud at a corner includes: Determine the distance between a point p in the unordered point cloud at the corner and its nearest neighbor points, and take the points whose distance is less than the radius threshold as the threshold point set; Calculate the covariance matrix of the threshold point set, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; Based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, the flattening index and normal of point p are determined as features of point p.

6. The segment corner reconstruction method based on scan data according to claim 5, characterized in that, The process of performing breadth-first clustering on point p includes: Sort all points p according to the flatness index; The sorted point sets are placed into an empty clustering container and numbered. Based on the normal of point p and all its neighboring points, cluster each point p corresponding to a given number. After clustering is complete, output the clustering container.

7. The segment corner reconstruction method based on scan data according to claim 6, characterized in that, For any two planar point clouds Pa and Pb, the surface reconstruction process includes: S101. Based on the planar distance threshold, obtain the optimal planes ha and hb corresponding to the planar point clouds Pa and Pb according to the RANSAC algorithm; S102. Determine the intersection line lab of the optimal planes ha and hb; S103. Project the planar point clouds Pa and Pb using the columnar filtering range, calculate the distance from the projected points to the intersection line lab, and retain the points projected within the columnar filtering range. S104. Determine the optimal plane corresponding to the retained points, and calculate the intersection line between the optimal planes; S105. Change the columnar filtering range and repeat S103 and S104 to obtain several sets of projection intersection results. S106. Select the optimal intersection line from several sets of projection intersection line results, and obtain the corner points under surface fitting and reconstruction based on the optimal intersection line.

8. The segment corner reconstruction method based on scan data according to claim 7, characterized in that, The method for determining corner points under surface fitting and reconstruction is as follows: After filtering out the optimal intersection lines using the normal distribution and applying a weighted average, k-clustering is performed. The corner points under surface fitting and reconstruction are determined based on the distance between the cluster center and the surrounding original point cloud cluster centers; The optimal intersection line is determined by clustering several sets of projected intersection line results; or... The ratio of the interior points contained in the two planes corresponding to the projection intersection line is determined.

9. A segment corner reconstruction system based on scan data, characterized in that, For performing the method according to any one of claims 1-8, comprising: The data import module is used to read the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud; The edge fitting and reconstruction module is used to perform edge fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under edge fitting and reconstruction. The surface fitting and reconstruction module is used to perform surface fitting and reconstruction based on the 3D point cloud of the bridge segment and the corresponding surface model of the point cloud, and output the corner points under surface fitting and reconstruction. The bridge segment corner point determination module is used to determine the corner points of bridge segments based on the corner points under edge fitting and reconstruction and the corner points under surface fitting and reconstruction.

Images