Method for extracting coordinates of bolt holes of splice plate of steel truss girder

Abstract

The invention belongs to the technical field of building construction, and discloses a steel truss splice plate bolt hole coordinate extraction method comprising the following steps: S1, collecting three-dimensional point cloud data of a splice plate; s2, carrying out data segmentation on the three-dimensional point cloud data, reserving a to-be-extracted bolt hole region, and carrying out fitting at the centroid of the to-be-extracted bolt hole region to generate reference frame line point cloud data with a preset size; s3, clustering the three-dimensional point cloud data of the to-be-extracted bolt hole area, S4, intercepting a two-dimensional image of a certain view in the three-dimensional point cloud data of the bolt hole after down-sampling, and carrying out graying and binarization processing; s5, performing perspective transformation on the binarized bolt hole two-dimensional data to obtain a bolt hole front view; and S6, on the basis of the bolt hole front view after perspective transformation and the three-dimensional coordinate information and the two-dimensional coordinate information of the four angular points in the reference frame line point cloud data, constructing a coordinate transformation matrix from two dimensions to three dimensions, and generating three-dimensional coordinates of each bolt hole, thereby reducing point cloud density and noise reduction while retaining the characteristics of the bolt holes.

Description

Technical Field

[0001] The present invention belongs to the technical field of building construction, and particularly relates to a method for extracting the coordinates of bolt holes in a steel truss girder splicing plate, which is used to repair the errors generated in the scanning process and data processing process of steel truss girder components and bolt hole groups of splicing plates, repair data gaps, and improve the subsequent assembly accuracy. Background Art

[0002] During the process of piecemeal assembly of steel truss girder segments of large steel truss girder bridges, pre-assembly is usually required to ensure the assembly matching performance of the girder segments and improve the installation quality of the girder segments. During the assembly process, first, the main truss members need to be precisely aligned through splicing plates, and then temporary drift pins are used for fixation to ensure the preliminary connection between the members. Next, the construction team will gradually increase the number of drift pins to enhance the connection stability. Subsequently, these drift pins will be gradually replaced with pre-designed high-strength bolts. Therefore, ensuring the accuracy of bolt holes is very important for the assembly of steel truss girders.

[0003] However, in the actual construction process, due to possible dimensional errors during the steel processing and transportation processes, some bolt holes may not be completely aligned with the holes on the splicing plates, resulting in the need for reaming or adjustment by construction workers during assembly. This not only increases the work difficulty but may also affect the stress distribution of the plate members. Inappropriate stress distribution will increase the fatigue of the structure and reduce the safety and lifespan of the bridge. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for extracting the coordinates of bolt holes in a steel truss girder splicing plate, which performs point cloud data processing, converts a three-dimensional image into a two-dimensional image analysis dataset, obtains the two-dimensional coordinates of bolt holes and then converts them back to three-dimensional coordinates, generates bolt hole point clouds for the clustered and downsampled point cloud data, and reduces the point cloud density and noise while retaining the bolt hole features.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: A method for extracting the coordinates of bolt holes in a steel truss girder splicing plate, the method comprising: S1: Collect the three-dimensional point cloud data of the splicing plate; S2: Perform data segmentation on the three-dimensional point cloud data, retain the bolt hole area to be extracted, and fit and generate reference frame line point cloud data of a preset size at the centroid of the bolt hole area to be extracted; S3: Cluster the three-dimensional point cloud data of the bolt hole area to be extracted to obtain the three-dimensional point cloud data of the bolt holes themselves and the shadow data generated due to light and shadow changes, remove the shadow data, and downsample the three-dimensional point cloud data of the bolt holes themselves; S4: Crop a 2D image of a certain view from the 3D point cloud data of the bolt hole itself after downsampling, perform grayscale conversion and binarization processing to obtain the binarized 2D data of the bolt hole; S5: Perform perspective transformation on the binarized 2D data of the bolt hole to obtain the front view of the bolt hole; S6: Based on the front view of the bolt hole after perspective transformation, and the 3D coordinate information and 2D coordinate information of the four corner points in the reference frame line point cloud data, construct a coordinate conversion matrix from 2D to 3D to generate the 3D coordinates of each bolt hole.

[0006] Further, the method further includes: S7: Generate a point cloud ring within a given radius range for the 3D point cloud data of the bolt hole itself after downsampling according to the 3D coordinates of the bolt hole group generated in S6 to enhance the bolt hole features.

[0007] Further, in S2, the preset size is larger than the area of the bolt hole to be extracted, and the 3D coordinates of the four corner points in the reference frame line point cloud data of the preset size are obtained.

[0008] Further, in S3, clustering is performed on the 3D point cloud data of the area of the bolt hole to be extracted. Specifically: Specify the search radius and the minimum point density threshold of the splicing plate data; Count the number of points within the search radius of the random seed points in the 3D point cloud data of the area of the bolt hole to be extracted, and use it as the density value of the corresponding seed points; Take each point in the 3D point cloud data of the area of the bolt hole to be extracted as a seed point, corresponding to the density value of each seed point; Take the seed points whose density values are greater than the minimum point density threshold of the splicing plate data as the 3D point cloud data of the splicing plate itself.

[0009] Further, in S3, downsampling is performed on the 3D point cloud data of the splicing plate itself. Specifically: Perform voxel filtering and Gaussian filtering on the 3D point cloud data of the splicing plate itself for downsampling.

[0010] Further, in S5, performing perspective transformation on the binarized 2D data of the bolt hole to obtain the front view of the bolt hole. Specifically: Identify the coordinates of the four corner points in the reference frame line point cloud data in the image as the source points pts1 for perspective transformation. Define the target points pts2 according to the set side length of the target image as 600 pixel points, and calculate the perspective transformation matrix according to the source points and the target points; Take the binarized 2D data of the bolt hole as the source image, and obtain the front view of the bolt hole after perspective transformation according to the source image and the perspective transformation matrix.

[0011] Further, S6 is specifically as follows: Based on the front view of the bolt holes after perspective transformation, use the Hough transform to detect circles, identify the diameters and center coordinates of each bolt hole, and construct a coordinate transformation matrix from 2D to 3D for the bolt holes based on the 3D and 2D coordinate information of the four corner points in the reference frame line point cloud data, and generate the 3D coordinates of each bolt hole.

[0012] An accurate extraction method for the positions of bolt holes on a steel truss girder splicing plate according to the present invention sequentially includes three-dimensional laser on-site scanning to obtain point cloud data, point cloud segmentation, generating standard-size frame line point cloud, intercepting images, clustering, downsampling, performing grayscale processing, threshold segmentation, perspective transformation, Hough circle detection, etc. on the intercepted images to obtain the bolt hole group and the two-dimensional coordinates of the frame line, constructing a two-dimensional to three-dimensional coordinate system, converting to obtain the three-dimensional coordinates of the bolt hole group, and generating bolt hole point cloud from the clustered and downsampled point cloud data to enhance the bolt hole features. The present invention processes the point cloud data, converts the three-dimensional image into a two-dimensional image analysis data set, converts back to the three-dimensional coordinates after obtaining the two-dimensional coordinates of the bolt holes, generates bolt hole point cloud from the clustered and downsampled point cloud data, and retains the bolt hole features while reducing the point cloud density and noise reduction. Description of the Drawings

[0013] Figure 1 It is a schematic diagram of the three-dimensional laser scanning process for obtaining point cloud data in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 2 It is a schematic diagram of segmenting the bolt hole group point cloud and fitting to generate a standard-size frame line in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 3 It is a schematic diagram of preprocessing (clustering) the point cloud data in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 4 It is a schematic diagram of preprocessing (downsampling) the point cloud data in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 5 It is a schematic diagram of performing grayscale processing on a two-dimensional image in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 6 It is a schematic diagram of performing binary processing on a two-dimensional image in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 7 It is a schematic diagram of performing perspective transformation and Hough circle detection on a two-dimensional image in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 8 It is a schematic diagram of enhancing the features of the preprocessed point cloud in a method for extracting the coordinates of bolt holes on a steel truss girder splicing plate in an embodiment; Figure 9 Schematic flow chart of a method for extracting bolt hole coordinates of a steel truss girder splicing plate in an embodiment. Specific embodiments

[0014] In the prior art, the pre-assembly of 1+1 truss segments of a steel truss girder bridge has high requirements for bolt hole accuracy, consumes a large amount of process resources, the preliminary preparation work required for image recognition is complex, and it is easily affected by on-site conditions. Point cloud recognition cannot directly extract the feature dot matrix information of bolt holes, and information loss will occur after preprocessing. Therefore, an embodiment of the present invention proposes a method for extracting bolt hole coordinates of a steel truss girder splicing plate.

[0015] The technical solution of the present invention will be further explained below by taking the splicing plate as an example in combination with the accompanying drawings and embodiments, but the present invention is not limited to the embodiments described below.

[0016] Perform laser scanning on the steel truss girder splicing plate to obtain the three-dimensional point cloud data of the splicing plate, as Figure 1 shown.

[0017] Segment the three-dimensional point cloud data of the splicing plate, retain the bolt hole area to be extracted, and fit and generate the reference frame line point cloud data with a size of 40 cm × 40 cm at the centroid of the bolt hole area to be extracted, Figure 2 shown in red in (the size of 40 cm × 40 cm is larger than the bolt hole area to be extracted).

[0018] Perform clustering processing on the three-dimensional point cloud data of the splicing plate using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method, cluster the three-dimensional point cloud data into the data of the splicing plate itself and the shadow data generated due to light and shadow changes, and discard the shadow data.

[0019] Specify the search radius radius and the minimum point density threshold minPts of the splicing plate data. Use the radiusSearch function to count the number of local points within radius of the random seed points of the three-dimensional point cloud as the point density value densities. Traverse the density of each point, and put the points that meet the density threshold into the point set. Finally, obtain the data clustering cluster of the splicing plate itself where the point cloud meets the threshold condition density.

[0020] As Figure 3 shown, the internal shadow area of the bolt hole is distinguished in gray.

[0021] Voxel filtering and Gaussian filtering are performed on the clustered spliced panel point cloud for downsampling. During downsampling, the standard deviation (sigma) is adjusted to control the width of the Gaussian function, maintaining a balance between the smoothing effect and detail retention; the sampling radius (radius) and voxel size (voxel_size) are adjusted to set the range of neighboring points and the point cloud density. The downsampled point cloud is as Figure 4 shown.

[0022] Intercept the 2D image of a certain view in the 3D point cloud data; Preprocess the intercepted 2D image. Perform grayscale processing on the picture through the cv2.cvtColor function to obtain a grayscale image, as Figure 5 shown, and then perform Gaussian blur processing on the grayscale image through the cv2.GaussianBlur function and binary processing through the cv2.threshold function, as Figure 6 shown.

[0023] Identify the coordinates of the four frame line corner points in the image through the defined pers_img_squar(img, src_list, length) as the source points pts1 of the perspective transformation. Define the target points pts2 according to the target image side length length = 600 dpi, and calculate the perspective transformation matrix based on the source points and target points.

[0024] According to the source image and the perspective transformation matrix, perform perspective transformation on the source image img to obtain the front view of the perspective-transformed image.

[0025] In the perspective transformation, the cv2.getPerspectiveTransform function and cv2.warpPerspective in openCV are mainly applied. The cv2.getPerspectiveTransform function accepts two parameters: pts1: Four points in the source image (the four vertices of the quadrilateral). pts2: Four points in the target image (usually the four vertices of the rectangle). Both of these parameters are 4x2 matrices, where each row is the (x, y) coordinates of a point. The function returns a 3x3 transformation matrix. The cv2.warpPerspective function accepts three parameters: img: The source image such as Figure 2 . matrix: The 3x3 transformation matrix calculated by cv2.getPerspectiveTransform, dsize: The size of the target image, usually the same as the size of the original image.

[0026] After obtaining the front view of the perspective transformation, detect circles through the Hough transform cv2.HoughCircles and output the 2D center coordinates, asFigure 7 As shown. The two-dimensional coordinates of the bolt hole centers obtained by Hough circle detection cannot be directly converted into three-dimensional coordinates. However, since all bolt holes are on the same plane, we can first assume that all two-dimensional coordinates (x, y) are on the plane (x, y, 0). Therefore, the coordinates of the target frame after perspective transformation are (0, 0, 0), (0, 600, 0), (600, 600, 0), and (600, 0, 0) in sequence.

[0027] Given the coordinates of the corner points of the generated reference frame lines, the centroids (i.e., the average values of all points) of the two-dimensional coordinate point set and the three-dimensional coordinate point set after dimension elevation can be calculated. Decentralize the two-dimensional coordinate point set and the three-dimensional coordinate point set after dimension elevation, that is, subtract their respective centroids. Construct the covariance matrix: Construct the covariance matrix matS through the two-dimensional coordinate point set and the three-dimensional coordinate point set after dimension elevation. SVD decomposition: Perform singular value decomposition (SVD) on the covariance matrix matS to obtain matS = UΣVT. Construct the rotation matrix: According to the results of SVD, construct the rotation matrix matR. Here, it is necessary to ensure that the determinant of the rotation matrix is 1 to ensure that it is an orthogonal matrix. If the determinant is -1, the last column of matM needs to be adjusted to make its determinant 1. Calculate the translation vector: Align the centroid of the two-dimensional coordinate point set after dimension elevation to the centroid of the three-dimensional coordinate point set through the rotation matrix and the translation vector. Combine rotation and translation: Combine the rotation matrix and the translation vector into a 4×4 transformation matrix T. That is, the rigid body transformation matrix from "two-dimensional" to "three-dimensional" is as follows: [[5.74299047e-02 8.19745597e-01 5.69841173e-01 7.60985410e-01] [-8.22210722e-02 5.72727820e-01 -8.15611758e-01 -1.32946121e+00] [-9.94958040e-01 -1.24466527e-05 1.00292066e-01 -1.01060912e-01] [0.00000000e+00 0.0000000e+00 0.00000000e+00 1.00000000e+00]] According to the obtained rigid body transformation matrix, the coordinates of the recognized bolt holes can be converted into the bolt hole coordinates in the three-dimensional point cloud space.

[0028] According to the three-dimensional coordinates of the bolt hole group centers and the standard radius obtained, generate a circular ring point cloud with 1e+4 points within the range of (radius R, radius R + 10mm) at the center, as follows Figure 8As shown, the red area is the generated point cloud, and the bolt hole features can be enhanced and standardized at the coordinates of the three-dimensional point cloud obtained by scanning.

[0029] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for extracting the bolt hole coordinates of a steel truss girder splicing plate, characterized in that The method comprises: S1: Collect 3D point cloud data of the spliced ​​plate; S2: segmenting the three-dimensional point cloud data, retaining the bolt hole area to be extracted, and fitting the centroid of the bolt hole area to be extracted to generate reference frame point cloud data of a preset size; S3: clustering the 3D point cloud data of the bolt hole area to be extracted, obtaining the 3D point cloud data of the bolt hole itself and the shadow data generated by light and shadow changes, removing the shadow data, and downsampling the 3D point cloud data of the bolt hole itself; S4: intercepting a 2D image of a certain view in the 3D point cloud data of the bolt hole after downsampling, performing grayscale and binarization processing, and obtaining binary 2D data of the bolt hole; S5: Perform perspective transformation on the binary bolt hole two-dimensional data to obtain a bolt hole front view; S6, based on the perspective transformed front view of the bolt hole, and the three-dimensional coordinate information and two-dimensional coordinate information of the four corner points in the reference frame point cloud data, construct a coordinate transformation matrix from two-dimensional to three-dimensional, and generate the three-dimensional coordinates of each bolt hole.

2. The method for extracting the bolt hole coordinates of the splicing plate of the steel truss girder according to claim 1, characterized in that, The method further comprises: S7: Generate a point cloud ring with a given radius range from the downsampled three-dimensional point cloud data of the bolt hole itself according to the three-dimensional coordinates of the bolt hole group generated in S6 to enhance the bolt hole feature.

3. A method for extracting the bolt hole coordinates of a steel truss girder splicing plate according to claim 1, characterized in that In S2, the preset size is larger than the bolt hole area to be extracted, and the three-dimensional coordinates of the four corner points in the reference frame point cloud data of the preset size are obtained.

4. The method for extracting the bolt hole coordinates of the splicing plate of the steel truss girder according to claim 3, wherein, In S3, the three-dimensional point cloud data of the bolt hole area to be extracted is clustered, specifically: Specify the search radius and the minimum point density threshold for the patch panel data; Count the number of points within the search radius of random seed points in the three-dimensional point cloud data of the bolt hole area to be extracted, and use it as the density value of the corresponding seed point; Each point in the three-dimensional point cloud data of the bolt hole area to be extracted is used as a seed point, and a density value corresponding to each seed point is obtained; The seed points whose density values ​​are greater than the minimum point density threshold of the splicing plate data are used as the three-dimensional point cloud data of the splicing plate itself.

5. A method for extracting the bolt hole coordinates of a steel truss girder splicing plate according to claim 4, characterized in that, In S3, the 3D point cloud data of the splicing plate itself is downsampled, specifically: The three-dimensional point cloud data of the splicing plate itself is downsampled by voxel filtering and Gaussian filtering.

6. The extraction method of the bolt hole coordinates of the steel truss girder splicing plate according to claim 4, characterized in that, In S5, the perspective transformation is performed on the binary bolt hole two-dimensional data to obtain the bolt hole front view, specifically: Identify the coordinates of the four corner points in the reference frame point cloud data in the image as the source point pts1 of the perspective transformation, define the target point pts2 according to the set target image side length of 600 pixels, and calculate the perspective transformation matrix based on the source point and the target point; The binary two-dimensional data of the bolt hole is used as a source image, and a perspective-transformed front view of the bolt hole is obtained according to the source image and a perspective transformation matrix.

7. A method for extracting the bolt hole coordinates of the splicing plate of a steel truss girder according to claim 6, characterized in that, S6 is specifically: Based on the front view of the bolt hole after perspective transformation, the Hough transform is used to detect the circle and identify the diameter and center coordinates of each bolt hole. Based on the three-dimensional coordinate information and two-dimensional coordinate information of the four corner points in the reference frame point cloud data, the coordinate transformation matrix of the bolt hole from two-dimensional to three-dimensional is constructed to generate the three-dimensional coordinates of each bolt hole.

Images