Steel truss girder chord member end bolt hole group information intelligent extraction method and system

Through the combination of random sampling consistency algorithm and point cloud normal vector, the automatic extraction of bolt hole group information at the end of the steel truss chord is achieved, solving the problems of low extraction efficiency and dependence on BIM model in the existing technology, and improving the digital pre-pre- Assembling efficiency and precision.

CN118941715APending Publication Date: 2024-11-12CHINA RAILWAY MAJOR BRIDGE ENG GRP CO LTD +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410998366.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing technology lacks high-precision, automation and intelligent methods to extract bolt hole group information at the end of the steel truss chord, resulting in low digital preassembly efficiency, and is sensitive to the initial value of the bolt hole group information, relying on BIM model design information, The bolt hole group information of multiple flat panels cannot be automatically extracted.

Method used

Using a random sampling consistency algorithm, multiple flat panels at the end of the chord are segmented and extracted by setting a distance threshold, curvature extraction and projection are performed by combining point cloud normal vectors, and circular contours are fitted using clustering and filtering algorithms to extract bolt holes. Group coordinates and radius information.

Benefits of technology

It realizes automatic, efficient and accurate extraction of bolt hole group information at the end of the steel truss chord, replacing the traditional cumbersome manual extraction, improving the digital preassembly efficiency, and solving the problems of initial value sensitivity and dependence on the BIM model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941715A_ABST
    Figure CN118941715A_ABST
Patent Text Reader

Abstract

The invention provides a method and a system for intelligently extracting information of bolt hole groups at the end part of a steel truss girder chord member, and the method comprises the steps: carrying out the segmentation extraction of a plurality of surface plates containing all bolt hole groups at the end part of the steel truss girder chord member through setting a distance threshold value by adopting a random sampling consistency algorithm; in combination with the point cloud normal vector, curvature extraction is carried out on the segmented and extracted surface plate point cloud; according to the extracted curvature, calculating a projection plane of the plane plate point cloud, and performing projection; clustering and filtering point clouds on a projection plane, fitting a circular contour by adopting a random sampling consistency algorithm, and extracting coordinates and radius information of bolt hole groups on all plane plates; according to the method, the coordinate and radius information of the bolt hole groups of all the surface plates at the end of the steel truss chord member can be automatically, efficiently and accurately extracted, manual extraction is replaced, meanwhile, according to the method, only original point cloud data need to be input, other design information and the like do not need to serve as initial conditions, and the digital pre-assembling efficiency of the large-section steel truss girder is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent bridge construction, and in particular to a method and system for intelligently extracting information of bolt hole groups at the ends of steel truss chords. Background Art

[0002] With the industrialization and green development of bridge manufacturing engineering, the assembly rate of bridge prefabrication is getting higher and higher. Among them, the digital pre-assembly technology, with its advantages of high precision, high efficiency, low carbonization and low cost, has gradually replaced the traditional physical pre-assembly technology of bridge components. In steel truss bridges with high requirements for bridge line shape, digital pre-assembly technology has attracted much attention. Digital pre-assembly technology is a technology that uses digital technology and computer-aided means to perform high-precision virtual assembly of engineering components.

[0003] The current steel truss digital pre-assembly technology, the implementation of the overall process, includes the following steps:

[0004] First, a 3D laser scanner is used to obtain the 3D point cloud data of the entire segment of the steel truss. In reverse engineering, the point data set of the product appearance surface obtained by measuring instruments is called a point cloud.

[0005] Then, the point cloud bolt hole group information of the chord ends of adjacent segments is extracted;

[0006] Next, the bolt hole group information is aligned with the bolt hole group information on the splicing plates at the two ends to obtain the assembly error.

[0007] The core of the above steps is the extraction of bolt hole group information.

[0008] Steel Truss Chord End: In steel truss structures, the chord end refers to the end of the main load-bearing member that constitutes the upper or lower edge of the truss. It connects the truss to other structural elements and is the key part of the truss.

[0009] 3D Digital Measurement is a measurement technology that uses advanced technology and equipment to accurately quantify and capture the three-dimensional size and shape of an object and convert it into a digital model.

[0010] PolyWorks software is a standard point cloud engineering software, mainly used for 3D scanning, point cloud scanning, dimensional analysis, comparison to CAD, reverse engineering, design, inspection, digital manufacturing and medical applications.

[0011] SolidWorks software is the world's first 3D CAD system developed based on Windows. It is an easy-to-learn and easy-to-use design tool based on features, parameterization, and solid modeling developed on Windows.

[0012] Due to the current lack of algorithms and software for extracting the bolt hole group information at the chord end, technical personnel in this field need to use commercial software such as PolyWorks or SolidWorks and perform a large amount of manual interaction. The amount of point cloud is very large. In this case, the extraction of bolt hole group information has seriously restricted the efficiency of digital pre-assembly.

[0013] Therefore, in the actual operation process, how to extract the bolt hole group information at the end of the chord with high precision, automation and intelligence is a difficult problem that troubles technical personnel in this field and restricts the development of digital pre-assembly.

[0014] Traditional bolt hole group information extraction methods are mostly based on the least squares method to fit the circle.

[0015] For nonlinear least squares problems, such as circle fitting, the least squares method usually requires an iterative solution. This means that the least squares method starts with an initial guess of the parameters and then, through an iterative process, gradually improves the parameters to minimize the error.

[0016] If the problem is nonlinear, the least squares method may fall into a local minimum rather than a global minimum. At this time, the initial guess of the bolt hole group information parameters, referred to as the bolt hole group information initial value, will affect the final result. In practical applications, the final result is very sensitive to the bolt hole group information initial value. Therefore, some people use the design information as the initial value to extract the bolt hole group information.

[0017] However, the least squares method is very sensitive and dependent on the initial value of the bolt hole group information. Under the premise of uncertain bolt hole positions, the least squares method cannot be used.

[0018] Secondly, the quality of field data collection is uneven, the bolt hole walls are not fully scanned, and the inner walls of most bolt hole walls collected are incomplete, making it difficult to effectively extract bolt hole group information.

[0019] In addition, there are multiple plane plates at the end of the chord of the steel truss. The current methods for extracting bolt hole group information are mostly improvements on the algorithms for extracting single circular holes for bolts. The extraction of bolt hole group information of the entire plane plate mostly relies on the design information obtained from the BIM (Building Information Modeling) model, and it is also impossible to automatically extract the bolt hole group information of multiple plane plates at the end of the steel truss chord. Summary of the invention

[0020] The present application provides a method and system for intelligently extracting information on bolt hole groups at the ends of steel truss chords, which can solve the technical problems existing in the prior art: being very sensitive and dependent on the initial values ​​of the bolt hole group information, incomplete inner wall collection, dependence on the design information of the BIM model during extraction, and inability to automatically extract multiple plane plates.

[0021] In a first aspect, an embodiment of the present application provides a method for intelligently extracting information about bolt hole groups at the ends of steel truss chords, the method comprising the following steps:

[0022] The random sampling consensus algorithm is used to segment and extract multiple plane plates containing all the bolt hole groups at the ends of the steel truss chords by setting the distance threshold.

[0023] Combined with the point cloud normal vector, the curvature of the plane plate point cloud extracted by segmentation is extracted;

[0024] According to the extracted curvature, the projection plane of the plane plate point cloud is calculated and projected;

[0025] The point cloud on the projection plane is clustered and filtered, and the circular contour is fitted using a random sampling consensus algorithm to extract the bolt hole group coordinates and radius information of all the plane plates at the end of the steel truss chord.

[0026] In one embodiment, the random sampling consensus algorithm is used to segment and extract multiple plane plates including all bolt hole groups at the ends of the steel truss chords by setting a distance threshold, which specifically includes the following steps:

[0027] When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated;

[0028] The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

[0029] In one implementation, the distance threshold has a value range of: thickness of a single plane plate < distance threshold < distance between two parallel plane plates.

[0030] In one embodiment, the curvature extraction of the plane plate point cloud extracted by segmentation is performed in combination with the point cloud normal vector, and specifically comprises the following steps:

[0031] Find the local neighborhood points of the inline points and form the covariance matrix of the neighborhood point coordinates;

[0032] Perform SVD decomposition on the covariance matrix and use the eigenvector corresponding to the minimum eigenvalue as the normal vector;

[0033] The minimum eigenvalue divided by the sum of all eigenvalues ​​represents the curvature at that point.

[0034] In one embodiment, the step of calculating the projection plane of the plane plate point cloud according to the extracted curvature and performing the projection comprises the following steps:

[0035] Sort the input point cloud curvature from small to large; take the curvature of 1 / 4 of the point cloud in order from small to large, and obtain the point cloud index; according to the index, obtain the point cloud corresponding to the sorted curvature;

[0036] Combined with the corresponding point cloud, the random sampling consensus algorithm is used to obtain the maximum plane where the point cloud is located by indexing the point cloud; all the inline points near the maximum plane are vertically projected along the normal vector direction of the maximum plane.

[0037] In one implementation, clustering and filtering the point cloud on the projection plane specifically includes the following steps:

[0038] Input the projected point cloud, set the neighborhood and radius, and use the density-based clustering algorithm DBSCAN to obtain all potential circular hole areas for the point cloud projected onto the plane;

[0039] Perform cluster filtering on the point cloud of the potential circular hole area to extract the circular hole contour and remove outliers;

[0040] The maximum side length of the minimum convex hull of the filtered point cloud is calculated. When the maximum side length of the convex hull meets the preset value range, circular hole fitting is performed, and each potential circular hole area is fitted cyclically.

[0041] In one embodiment, the method of fitting a circular contour using a random sampling consensus algorithm specifically comprises the following steps:

[0042] Select any three non-collinear points in the point cloud after cluster filtering, fit the spatial circle equation, calculate the distance from each point cloud to the center of the circle - the radius of the circle according to the calculated coordinates and radius of the circle center, and select the point whose calculation result is less than the preset value as the circumference point;

[0043] In the loop iteration, the number of circumferential points changes randomly. When the number of iterations is greater than the second iteration threshold, the equation of the circle corresponding to the largest number of circumferential points is found, which is the fitting circle, and the data is saved.

[0044] In a second aspect, an embodiment of the present application provides an intelligent extraction system for information of bolt hole groups at the ends of steel truss chords, the intelligent extraction system comprising:

[0045] The plane plate segmentation module is used to segment and extract multiple plane plates including all bolt hole groups at the ends of steel truss chords by using a random sampling consensus algorithm and setting a distance threshold;

[0046] The curvature extraction module is used to extract the curvature of the plane plate point cloud extracted by segmentation in combination with the point cloud normal vector;

[0047] The projection module is used to: calculate the projection plane of the plane plate point cloud according to the extracted curvature and perform projection;

[0048] The clustering and filtering module is used to cluster and filter the point cloud on the projection plane;

[0049] The circular hole fitting module is used to: fit the circular contour using a random sampling consensus algorithm and extract the bolt hole group coordinates and radius information of all the plane plates at the end of the steel truss chord.

[0050] In one embodiment, the intelligent extraction system further comprises:

[0051] A preprocessing module is used to: preprocess the input chord end point cloud data;

[0052] The point cloud quantity judgment module is used to: start the plane plate segmentation module when it is determined that the point cloud quantity of the chord end after preprocessing is greater than the preset threshold value of the remaining point cloud quantity of the chord end.

[0053] In one embodiment, the planar panel segmentation module is specifically used for:

[0054] When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated;

[0055] The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

[0056] The beneficial effects brought by the technical solution provided by the embodiments of the present application include:

[0057] (1) In the random sampling consensus algorithm, the present application sets a distance threshold to segment and extract multiple plane plates including all bolt hole groups at the ends of the steel truss chords, thereby solving the problem of automatic segmentation of multiple plane plates;

[0058] (2) The present application combines the point cloud normal vector to extract the curvature of the plane plate point cloud extracted by segmentation; calculates the projection plane of the plane plate point cloud according to the extracted curvature and projects it; clusters and filters the point cloud on the projection plane to solve the technical problem of incomplete inner wall acquisition;

[0059] (3) This application uses a random sampling consensus algorithm to segment and extract multiple plane plates containing all bolt hole groups at the ends of steel truss chords, cluster and filter the point cloud on the projection plane, and solve the technical problem of relying on the design information of the BIM model during extraction;

[0060] (4) This application adopts a random sampling consensus algorithm, instead of the least squares method, to fit the circular contour, solving the technical problem of being very sensitive and dependent on the initial value of the bolt hole group information;

[0061] (5) Combining the above steps, the present application can automatically, efficiently and accurately extract the bolt hole group coordinates and radius information of all plane plates at the end of the steel truss chord, replacing the traditional cumbersome manual extraction. At the same time, this method only needs to input the original point cloud data, and does not require other design information as initial conditions, which effectively improves the efficiency of digital pre-assembly of large-segment steel trusses. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of an embodiment of a method for intelligently extracting information of bolt hole groups at the ends of steel truss chords of the present application;

[0063] Figure 2 This is a schematic diagram of the structure of the end of the steel truss chord of this application;

[0064] Figure 3 For this application Figure 1 A detailed flow chart of step 104;

[0065] Figure 4 For this application Figure 1 A detailed flow chart of step 105;

[0066] Figure 5 For this application Figure 1 A detailed flow chart of step 106;

[0067] Figure 6 For this application Figure 1 A detailed flow chart of step 107;

[0068] Figure 7For this application Figure 1 A detailed flow chart of step 108;

[0069] Figure 8 This is a schematic diagram of the architecture of an embodiment of a system for intelligently extracting information about bolt holes at the ends of steel truss chords of the present application;

[0070] Fig. 9 This is a schematic diagram of the architecture of another embodiment of the system for intelligently extracting information about bolt hole groups at the ends of steel truss chords of the present application.

[0071] In the figure: 201 - first plane plate, 202 - second plane plate, 203 - third plane plate, 204 - fourth plane plate. DETAILED DESCRIPTION

[0072] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0073] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0074] In a first aspect, an embodiment of the present application provides a method for intelligently extracting information of bolt hole groups at the ends of steel truss chords, comprising the following steps:

[0075] The random sampling consensus algorithm is used to segment and extract multiple plane plates containing all the bolt hole groups at the ends of the steel truss chords by setting the distance threshold.

[0076] Combined with the point cloud normal vector, the curvature of the plane plate point cloud extracted by segmentation is extracted;

[0077] According to the extracted curvature, the projection plane of the plane plate point cloud is calculated and projected;

[0078] The point cloud on the projection plane is clustered and filtered, and the circular contour is fitted using a random sampling consensus algorithm to extract the bolt hole group coordinates and radius information of all the plane plates at the end of the steel truss chord.

[0079] The above random sampling consensus algorithm is used to segment and extract multiple plane plates containing all bolt hole groups at the ends of the steel truss chords by setting a distance threshold, which specifically includes the following steps:

[0080] When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated;

[0081] The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

[0082] The value range of the above distance threshold is: thickness of a single plane plate < distance threshold < distance between two parallel plane plates.

[0083] The above method combines the point cloud normal vector to extract the curvature of the plane plate point cloud extracted by segmentation, which specifically includes the following steps:

[0084] Find the local neighborhood points of the inline points and form the covariance matrix of the neighborhood point coordinates;

[0085] Perform SVD decomposition on the covariance matrix and use the eigenvector corresponding to the minimum eigenvalue as the normal vector. The minimum eigenvalue divided by the sum of all eigenvalues ​​represents the curvature of the point.

[0086] The above method calculates the projection plane of the plane plate point cloud based on the extracted curvature and performs projection, which includes the following steps:

[0087] Sort the input point cloud curvature from small to large; take the curvature of 1 / 4 of the point cloud in order from small to large, and obtain the point cloud index; according to the index, obtain the point cloud corresponding to the sorted curvature;

[0088] Combined with the corresponding point cloud, the random sampling consensus algorithm is used to obtain the maximum plane where the point cloud is located by indexing the point cloud; all the inline points near the maximum plane are vertically projected along the normal vector direction of the maximum plane.

[0089] The above-mentioned clustering and filtering of the point cloud on the projection plane specifically includes the following steps:

[0090] Input the projected point cloud, set the neighborhood and radius, and use the density-based clustering algorithm DBSCAN to obtain all potential circular hole areas for the point cloud projected onto the plane;

[0091] Perform cluster filtering on the point cloud of the potential circular hole area to extract the circular hole contour and remove outliers;

[0092] The maximum side length of the minimum convex hull of the filtered point cloud is calculated. When the maximum side length of the convex hull meets the preset value range, circular hole fitting is performed, and each potential circular hole area is fitted cyclically.

[0093] The above-mentioned random sampling consensus algorithm is used to fit the circular contour, which specifically includes the following steps:

[0094] Select any three non-collinear points in the point cloud after cluster filtering, fit the spatial circle equation, calculate the distance from each point cloud to the center of the circle - the radius of the circle according to the calculated coordinates and radius of the circle center, and select the point whose calculation result is less than the preset value as the circumference point;

[0095] In the loop iteration, the number of circumferential points changes randomly. When the number of iterations is greater than the second iteration threshold, the equation of the circle corresponding to the largest number of circumferential points is found, which is the fitting circle, and the data is saved.

[0096] The following describes them in detail through specific embodiments.

[0097] In one embodiment, referring to Figure 1 As shown, Figure 1 This is a flow chart of the first embodiment of the method for intelligently extracting bolt hole group information from the end of a steel truss chord in the present application. The end of a chord of a steel truss is taken as an example for explanation.

[0098] like Figure 1 As shown, in this embodiment, the method for intelligently extracting information of bolt hole groups at the ends of steel truss chords includes the following steps:

[0099] Step 101: Input the point cloud data of the end of the steel truss chord, and go to step 102.

[0100] Generally, each chord end of a steel truss includes 4 plane plates and 4 stiffening ribs. Therefore, there are point cloud data of multiple plane plates to be processed. These plane plates already contain all the bolt hole groups at the ends of the steel truss chords. Each plane plate includes a main plate and two bottom surfaces. Figure 2 As shown, in this embodiment, there are 4 plane plates to be processed, specifically including Figure 2 As shown in FIG. 1 , the first plane plate 201 , the second plane plate 202 , the third plane plate 203 , and the fourth plane plate 204 , the order of the four plane plates to be processed is randomly arranged. In the actual processing process, the plane plates need to be processed one by one in sequence.

[0101] Step 102: pre-process the chord end point cloud data input in step 101, mainly adopting existing point cloud data pre-processing processes such as data clipping and data filtering, and then go to step 103.

[0102] The point cloud data preprocessing of the chord end includes: data cleaning, format conversion, and point cloud downsampling.

[0103] Data cleaning methods include: outlier detection and removal, and conversion of point cloud format to efficiently organized PCD format. Point Cloud PCD format (Point Cloud Data format) is a file format specifically used to store point cloud data. This format was introduced by the open source library Point Cloud Library (PCL) and is widely used in 3D scanning and 3D imaging technology. The PCD format is designed to store and read point cloud data quickly and efficiently, while supporting multiple data types and rich metadata. This format is faster to process.

[0104] Point cloud downsampling methods include voxel downsampling, random downsampling and so on.

[0105] For the existing 3D scanning point cloud data of the steel truss chord end, the large blocks of obvious outliers and outlier noise points are first roughly screened through point cloud clipping, and then the Z-Score algorithm is used to remove the outliers. Z-Score is a common method of data processing, which can convert data of different magnitudes into a unified Z score for comparison.

[0106] The processed point cloud is processed using voxel downsampling, and the point cloud with a radius within the set value is replaced by a point to achieve the downsampling effect. The radius value here is set within 0.1 to 0.3, preferably 0.15.

[0107] Step 103: Determine the number of point clouds:

[0108] Determine whether the number of point clouds at the end of the chord is greater than a preset threshold value of the number of remaining point clouds at the end of the chord. The numerical range of the threshold value of the number of remaining point clouds at the end of the chord can be preset to 3000-5000. In this embodiment, the threshold value of the number of remaining point clouds at the end of the chord is preferably 4000. If the number of point clouds at the end of the chord is greater than 4000, go to step 104; if the number of point clouds at the end of the chord is less than or equal to 4000, the process ends.

[0109] Since the plane plates are extracted one by one, after extracting a plane plate, the inline points of the plane plate are removed from the original point cloud, and then the remaining point cloud volume is calculated to see if it is necessary to continue extracting the plane plate. If the remaining point cloud volume is less than or equal to 4000 (60,000 to 100,000 of the total point volume, and the total point volume in actual operation is approximately 6 million point clouds), the process ends; if it is greater than 4000, proceed to the next step.

[0110] Step 104: Using the RANSAC (RANdom SAmple Consensus) algorithm, based on the equation for segmenting the plane plate and combined with the spatial distance, the inline points contained in each plane plate are obtained, and the largest plane of the multi-plane plate point cloud data is extracted, that is, the extracted plane plate is segmented, and the inline points in the input point cloud data are removed, and then go to step 105.

[0111] Step 105: Calculate the curvature of the inline point: for each inline point extracted from the plane plate, calculate the normal vector of each point cloud; extract the curvature of the point cloud data according to the normal vector of the point cloud; go to step 106;

[0112] Step 106: Calculate the projection plane based on the curvature of the inline point, project the point cloud data of the inline point, and go to step 107;

[0113] Step 107: clustering and filtering the projected point cloud data, and then go to step 108;

[0114] Step 108: Use a random sampling consensus algorithm to fit the clustered and filtered circular hole point cloud, save the data, return to step 104, and refer to the process of step 104 to continue extraction until the extraction is completed.

[0115] Further, in one embodiment, see Figure 3 As shown, the above step 104 extracts the largest plane of the point cloud data of the multi-plane plates through a random sampling consensus algorithm; obtains the inline points contained in each plane plate according to the equation for splitting the plane plate and combining the spatial distance, and removes the inline points of the plane plate, which specifically includes the following steps:

[0116] Step 1041: input the voxel-downsampled point cloud plane of the end of the steel truss chord. The point cloud data at this step has been preprocessed, and the number of point clouds is greater than 4000, so go to step 1042;

[0117] Step 1042: randomly select three non-collinear points, fit a plane, and go to step 1043;

[0118] Step 1043: Calculate the spatial distance between the plane fitted in step 1042 and other point clouds, and go to step 1044;

[0119] Step 1044: Determine whether the spatial distance calculated in step 1043 is less than a distance threshold. Theoretically, the value range of the distance threshold that can be set is:

[0120] The thickness of a single plane plate < the set distance threshold < the distance between two parallel plane plates;

[0121] The distance threshold set in this embodiment is preferably 48, and 48 is only a value listed in the numerical range of the distance threshold that can be set in this embodiment.

[0122] If the spatial distance is less than 48, go to step 1046; otherwise, go to step 1045;

[0123] Step 1045: taking the point cloud whose spatial distance to the plane fitted in step 1042 is greater than or equal to 48 as the external point;

[0124] Step 1046: The point cloud whose spatial distance to the plane fitted in step 1042 is less than 48 is used as inline points. Here, the number of inline points is recorded as Ni, and then go to step 1047;

[0125] Step 1047: Determine whether the number of iterations is less than a first iteration threshold, the value range of the first iteration threshold is [1000, 10000], preferably [5000, 6000]. If yes, return to step 1042; otherwise, go to step 1048;

[0126] Step 1048: Each time an iteration is performed, new inline points are randomly generated, and the number of inline points Ni changes randomly. The maximum value of the number of inline points Ni is recorded as Nj. The plane corresponding to the iteration corresponding to Nj, that is, the plane corresponding to the iteration with the most inline points, is used as the final fitting plane. The final fitting plane is the plane plate extracted by segmentation. Then, the inline points in the input point cloud data are removed to avoid errors caused by repeated calculations.

[0127] Remaining point cloud processing: The point cloud data of the inline points of the largest plane plate will be eliminated, and the next largest plane plate extraction and circular hole fitting will be performed until the number of point clouds is less than 4000.

[0128] This embodiment uses the RANSAC algorithm to automatically segment the four plane plates containing all the bolt hole groups at the ends of the chords of the steel truss beam by setting a distance threshold, thereby solving the problem of automatically segmenting multiple plane plates into a single plane plate. Previously, multiple plane plates were usually cut out manually one by one based on commercial software such as PolyWorks software or SolidWorks software.

[0129] It should be noted that the distance threshold set in this embodiment is 48 mm, the number of random sampling points each time is 3, and the maximum number of iterations is not less than 1000 times, preferably 2000 times.

[0130] Further, in one embodiment, see Figure 4As shown, the above step 105, calculating the curvature of the inline point: for each inline point extracted from the plane plate, calculating the normal vector of each point cloud; extracting the curvature of the point cloud data according to the normal vector of the point cloud, specifically includes the following steps:

[0131] Step 1051: input the original point cloud, and go to step 1052;

[0132] Step 1052: Calculate the maximum plane and the inline points, and go to step 1053;

[0133] Step 1053: Find the local neighborhood points of the inline point, and go to step 1054;

[0134] Step 1054: construct the covariance matrix of the neighborhood point coordinates, and go to step 1055;

[0135] Step 1055: Decompose the covariance matrix by SVD (Singular Value Decomposition), use the minimum eigenvalue as the normal vector, and go to step 1056;

[0136] Step 1056: Calculate the angles of all normal vectors in the neighborhood. The variance of the angle change is the curvature. End.

[0137] Due to the influence of the thickness of the plane plate, if the maximum plane inline points are directly used for projection, the obtained plane may be an inclined plane. Because the number of point clouds contained in the inline points is too large, the fitted plane is not on the bottom surface of the chord, but on the inclined plane in the middle of the screw hole. If this plane is used, some screw holes may have projection deviations, or the projection may be an ellipse instead of a circle, affecting the extraction accuracy. Therefore, it is not advisable to directly perform plane projection based on the inline points extracted from the maximum plane.

[0138] For each point, first use the k-nearest neighbor search to find its neighborhood point set. The k-nearest neighbor (KNN) classification algorithm is a relatively mature method in theory and one of the simplest machine learning algorithms. The idea of ​​this method is: in the feature space, if most of the k nearest samples (i.e. the nearest ones in the feature space) near a sample belong to a certain category, then the sample also belongs to this category.

[0139] Then the covariance matrix of the neighborhood point set is calculated and principal component analysis is performed.

[0140] The eigenvector corresponding to the minimum eigenvalue is used as the normal vector of the point, and the minimum eigenvalue divided by the sum of all eigenvalues ​​represents the curvature of the point.

[0141] Further, in one embodiment, see Figure 5As shown, the above step 106: calculating the projection plane based on the curvature of the inline point and projecting the point cloud data of the inline point specifically includes the following steps:

[0142] Step 1061: input the point cloud curvature, and go to step 1062;

[0143] Step 1062: Sort the curvatures from small to large, and go to step 1063;

[0144] Step 1063: From small to large, take the curvature of 1 / 4 of the point cloud volume, obtain the point cloud index, and go to step 1064;

[0145] Step 1064: Obtain the point cloud corresponding to the sorted curvature according to the index, and go to step 1065;

[0146] Step 1065: According to the corresponding point cloud, use the random sampling consensus algorithm to obtain the maximum plane where the point cloud is located by indexing the point cloud, and go to step 1066;

[0147] Step 1066: Project all inline points near the maximum plane in step 1065 vertically along the normal vector direction of the maximum plane, and end.

[0148] According to the characteristic that the surface curvature of the bolt hole is much smaller than the curvature of the middle inner wall, the obtained curvatures are sorted from small to large, and the index position point cloud of the curvature of 1 / 4 of the number of inline points is used as the standard projection point cloud.

[0149] Then, a random sampling consensus algorithm is used to perform projection calculation on the point cloud. The distance threshold here is 0.01 mm, the number of random sampling points is 3, and the maximum number of iterations is 5000. The projection plane obtained at this time is the exact bottom surface of the inline point.

[0150] According to the acquired projection plane, the point cloud data of the inline point is projected along the normal direction of the plane, and finally the spatial point is projected onto the spatial plane.

[0151] Further, in one embodiment, see Figure 6 As shown, the above step 107: clustering and filtering the projected point cloud data specifically includes the following steps:

[0152] Step 1071: input the projected point cloud, and go to step 1072;

[0153] Step 1072: Set the neighborhood to 5 points and the radius to 0.5 mm, and go to step 1073;

[0154] Step 1073: For the point cloud projected onto the plane, use the density clustering method to obtain all potential circular hole areas, and the DBSCAN algorithm can be used for clustering to obtain all clusters, and then go to step 1074;

[0155] DBSCAN (Density-Based Spatial Clustering of Application with Noise) is a representative density-based clustering algorithm. Different from partitioning and hierarchical clustering algorithms, it defines a cluster as the largest set of density-connected points, divides areas with sufficiently high density into clusters, and can find clusters of arbitrary shapes in a noisy spatial database.

[0156] Step 1074: perform statistical outlier removal on the potential hole area point cloud, perform statistical outlier removal on the clustering result, with a removal radius of 5 mm and a standard deviation of 1 mm, and then go to step 1075;

[0157] Step 1075: Calculate the maximum side length of the minimum convex hull of the filtered point cloud, and go to step 1076;

[0158] Step 1076: After density clustering, the maximum convex hull is calculated for the potential circular hole position. If the maximum side length of the convex hull is too large or too small, the circular hole is not fitted. When 20 mm < maximum side length of the convex hull < 50 mm, circular hole fitting is performed. Each potential circular hole area is fitted cyclically. For the potential circular hole area after filtering, the least squares and random sampling consensus algorithms can be used to fit the circle, and the process ends.

[0159] The point cloud on the inline point plane, when projected, has a density much smaller than the projection of the point cloud on the inner wall of the bolt hole on the plane. The number of points within the radius is generally between 5 and 10. The density-based spatial clustering algorithm DBSCAN is used, with a neighborhood radius of 0.5 mm and a minimum number of sampling point clouds of 5. The point cloud is organized using a KD tree to speed up the algorithm process.

[0160] The clustered point cloud needs to be filtered because the above clustering algorithm will extract some areas with high edge density as potential screw hole areas. In actual operation, there are about 6 million point clouds, nearly 30,000 potential areas, and only 72 areas actually contain screw holes. The values ​​used in this embodiment are for reference only.

[0161] To avoid the situation where the potential point cloud cannot be fitted into a space circle, first use the minimum convex hull method to perform large-scale elimination: specifically, obtain the convex hull of each potential screw hole area, calculate the maximum side length of the convex hull, and if it is much larger than the diameter of the standard screw hole 33 mm, defined as greater than 50 mm here, or much smaller than 33 mm, defined as less than 5 mm here, the area will be deleted. The values ​​used in this embodiment are for reference only.

[0162] After filtering by the convex hull, density filtering is performed again. Gaussian density filtering is used to remove the point cloud of outliers to improve the accuracy of circular hole detection.

[0163] The remaining segmented planes are processed in the same manner as above.

[0164] This embodiment calculates the projection plane based on the curvature of the inline points, and combines the clustering and filtering methods of the projected point cloud data, which can work together to ultimately eliminate the problem of incomplete hole wall scanning.

[0165] Further, in one embodiment, see Figure 7 As shown, step 108: using a random sampling consensus algorithm to fit the clustered and filtered circular hole point cloud and save the data, specifically includes the following steps:

[0166] Step 1081: select any three non-collinear points in the point cloud after projection filtering, fit the space circle equation, and go to step 1082;

[0167] Step 1082: Calculate the distance from each point cloud to the center of the circle minus the radius of the circle according to the calculated coordinates of the center of the circle and the radius, and go to step 1083;

[0168] Step 1083: select the point whose calculation result in step 1082 is less than the preset value as the circumference point. The preset value can be selected independently according to the degree of data dispersion. For example, it can be set to 0.2 mm, and then go to step 1084;

[0169] Step 1084: determine whether the number of iterations is less than a second iteration threshold, the value range of the second iteration threshold is [1000, 10000], preferably [5000, 6000]. If yes, return to step 1081 and continue fitting; otherwise, go to step 1085;

[0170] Step 1085: Each time iterates, new circumferential points are randomly generated, and the number of circumferential points changes randomly. The cycle iterates and the equation of the circle corresponding to the circle with the largest number of circumferential points is found, which is the fitting circle. The data is saved and the process ends.

[0171] The potential area holes screened above are fitted with a plane circle using a random sampling consensus algorithm combined with an error loss function, and a multi-process approach is used. When the fitting time exceeds 7 seconds, it is automatically skipped and the next one is fitted, because a long fitting time often means that the potential area is unreasonable. Finally, the center coordinates and radius of all fitted circles are output to an Excel table for storage and named uniformly.

[0172] In this embodiment, a random sampling consensus algorithm is used instead of a least squares algorithm, thereby avoiding the selection of initial values.

[0173] In this embodiment, clustering is used to screen out potential areas containing screw holes from a large number of point clouds, filtering is used to remove scattered point clouds in the screw hole area, and the convex hull method is used to remove unreasonable areas. In this way, it is possible to ensure that higher algorithm processing efficiency is obtained without missing the fitting circular hole.

[0174] This embodiment adopts a random sampling consensus algorithm instead of the least squares method to fit the circle, which has a better effect on fitting a spatial plane circle and solves the problem of bolt extraction relying on the initial value. It can finally realize the intelligent extraction of the bolt hole group coordinates and radius information in the complex multi-plane chord end point cloud, replacing the traditional cumbersome manual extraction, effectively improving the efficiency of digital pre-assembly of large-segment steel trusses and reducing costs.

[0175] In a second aspect, an embodiment of the present application also provides a system for intelligently extracting information about bolt hole groups at the ends of steel truss chords.

[0176] In one embodiment, referring to Figure 8 , Figure 8 This is a schematic diagram of the structure of an embodiment of the intelligent extraction system of bolt hole group information at the end of the steel truss chord of the present application. Figure 8 As shown, the intelligent extraction system of bolt hole group information at the end of the steel truss chord includes:

[0177] The plane plate segmentation module is used to segment and extract multiple plane plates including all bolt hole groups at the ends of steel truss chords by using a random sampling consensus algorithm and setting a distance threshold;

[0178] The curvature extraction module is used to extract the curvature of the plane plate point cloud extracted by segmentation in combination with the point cloud normal vector;

[0179] The projection module is used to: calculate the projection plane of the plane plate point cloud according to the extracted curvature and perform projection;

[0180] The clustering and filtering module is used to cluster and filter the point cloud on the projection plane;

[0181] The circular hole fitting module is used to: fit the circular contour using a random sampling consensus algorithm and extract the bolt hole group coordinates and radius information of all the plane plates at the end of the steel truss chord.

[0182] Further, in one embodiment, referring to Fig. 9 , Fig. 9 This is a schematic diagram of another embodiment of the intelligent extraction system for bolt hole group information at the end of a steel truss chord of the present application. Fig. 9As shown, the intelligent extraction system for bolt hole group information at the end of the steel truss chord also includes:

[0183] The preprocessing module is used to preprocess the input point cloud data of the chord end, mainly adopting existing point cloud data preprocessing processes such as data cropping and data filtering. The preprocessing of the point cloud data of the chord end includes: data cleaning, format conversion, and point cloud downsampling.

[0184] Further, in one embodiment, referring to Fig. 9 , Fig. 9 This is a schematic diagram of another embodiment of the intelligent extraction system for bolt hole group information at the end of a steel truss chord of the present application. Fig. 9 As shown, the intelligent extraction system for bolt hole group information at the end of the steel truss chord also includes:

[0185] The point cloud quantity judgment module is used to judge whether the number of point clouds at the end of the chord after preprocessing is greater than the preset threshold value of the number of remaining point clouds at the end of the chord. The numerical range of the threshold value of the number of remaining point clouds at the end of the chord can be preset to 3000-5000. In this embodiment, the threshold value of the number of remaining point clouds at the end of the chord is preferably 4000. If the number of point clouds at the end of the chord is greater than 4000, the plane plate segmentation module is started; if the number of point clouds at the end of the chord is less than or equal to 4000, the process ends.

[0186] Since the plane plates are extracted one by one, after a plane plate is extracted, the inline points of the plane plate are removed from the original point cloud, and then the remaining point cloud volume is calculated to see whether it is necessary to continue extracting the plane plate. If the remaining point cloud volume is less than or equal to 4000 (60,000 to 100,000 of the total point volume, and the total point volume in actual operation is approximately 6 million point clouds), the process ends; if it is greater than 4000, the plane plate segmentation module is started.

[0187] Plane panel splitting module, specifically used for:

[0188] When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated;

[0189] The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

[0190] Among them, the functional implementation of each module in the above-mentioned intelligent extraction system of bolt hole group information at the end of the steel truss chord corresponds to the steps in the above-mentioned embodiment of the intelligent extraction method of bolt hole group information at the end of the steel truss chord, and its functions and implementation processes will not be repeated here one by one.

[0191] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0192] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.

[0193] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.

[0194] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; the “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0195] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0196] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for intelligently extracting information about bolt holes at the ends of steel truss chords, characterized in that: The intelligent extraction method comprises the following steps: The random sampling consensus algorithm is used to segment and extract multiple plane plates containing all the bolt hole groups at the ends of the steel truss chords by setting the distance threshold. Combined with the point cloud normal vector, the curvature of the plane plate point cloud extracted by segmentation is extracted; According to the extracted curvature, the projection plane of the plane plate point cloud is calculated and projected; The point cloud on the projection plane is clustered and filtered, and the circular contour is fitted using a random sampling consensus algorithm to extract the coordinates and radius information of the bolt hole groups on all the plane plates at the ends of the steel truss chords.

2. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1, characterized in that: The random sampling consensus algorithm is used to segment and extract multiple plane plates containing all bolt hole groups at the ends of the steel truss chords by setting a distance threshold, which specifically includes the following steps: When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated; The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

3. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1 or 2, characterized in that: The value range of the distance threshold is: thickness of a single plane plate < distance threshold < distance between two parallel plane plates.

4. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1, characterized in that: The method of extracting curvature of the plane plate point cloud extracted by segmentation by combining the point cloud normal vector specifically comprises the following steps: Find the local neighborhood points of the inline points and form the covariance matrix of the neighborhood point coordinates; Perform SVD decomposition on the covariance matrix and use the eigenvector corresponding to the minimum eigenvalue as the normal vector; The minimum eigenvalue divided by the sum of all eigenvalues ​​represents the curvature at that point.

5. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1, characterized in that: The method of calculating the projection plane of the plane plate point cloud according to the extracted curvature and performing projection comprises the following steps: Sort the input point cloud curvature from small to large; take the curvature of 1 / 4 of the point cloud in order from small to large, and obtain the point cloud index; according to the index, obtain the point cloud corresponding to the sorted curvature; Combined with the corresponding point cloud, the random sampling consensus algorithm is used to obtain the maximum plane where the point cloud is located by indexing the point cloud; all the inline points near the maximum plane are vertically projected along the normal vector direction of the maximum plane.

6. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1, characterized in that: The clustering and filtering of the point cloud on the projection plane specifically includes the following steps: Input the projected point cloud, set the neighborhood and radius, and use the density-based clustering algorithm DBSCAN to obtain all potential circular hole areas for the point cloud projected onto the plane; Perform cluster filtering on the point cloud of the potential circular hole area to extract the circular hole contour and remove outliers; The maximum side length of the minimum convex hull of the filtered point cloud is calculated. When the maximum side length of the convex hull meets the preset value range, circular hole fitting is performed, and each potential circular hole area is fitted cyclically.

7. The method for intelligently extracting information of bolt hole groups at the ends of steel truss chords according to claim 1, characterized in that: The method of fitting a circular contour by using a random sampling consensus algorithm specifically comprises the following steps: Select any three non-collinear points in the point cloud after cluster filtering, fit the spatial circle equation, calculate the distance from each point cloud to the center of the circle - the radius of the circle according to the calculated coordinates and radius of the circle center, and select the point whose calculation result is less than the preset value as the circumference point; In the loop iteration, the number of circumferential points changes randomly. When the number of iterations is greater than the second iteration threshold, the equation of the circle corresponding to the largest number of circumferential points is found, which is the fitting circle, and the data is saved.

8. An intelligent extraction system for bolt hole group information at the end of a steel truss chord, characterized in that: The intelligent extraction system comprises: The plane plate segmentation module is used to segment and extract multiple plane plates including all bolt hole groups at the ends of steel truss chords by using a random sampling consensus algorithm and setting a distance threshold; The curvature extraction module is used to extract the curvature of the plane plate point cloud extracted by segmentation in combination with the point cloud normal vector; The projection module is used to: calculate the projection plane of the plane plate point cloud according to the extracted curvature and perform projection; The clustering and filtering module is used to cluster and filter the point cloud on the projection plane; The circular hole fitting module is used to fit the circular contour using a random sampling consensus algorithm and extract the coordinates and radius information of the bolt hole groups on all the plane plates at the ends of the steel truss chords.

9. The intelligent extraction system for bolt hole group information at the end of a steel truss chord according to claim 8, characterized in that: The intelligent extraction system also includes: A preprocessing module is used to: preprocess the input chord end point cloud data; The point cloud quantity judgment module is used to: start the plane plate segmentation module when it is determined that the point cloud quantity of the chord end after preprocessing is greater than the preset threshold value of the remaining point cloud quantity of the chord end.

10. The intelligent extraction system for bolt hole group information at the end of a steel truss chord according to claim 8 or 9, characterized in that: The plane panel segmentation module is specifically used for: When the number of point clouds at the end of the chord is greater than the preset threshold of the number of remaining point clouds at the end of the chord, three non-collinear points are randomly selected from the pre-processed point cloud data, a plane is fitted, and the spatial distance between the fitted plane and other point clouds is calculated; The point clouds whose spatial distance to the fitted plane is less than the set distance threshold are taken as inline points, and the number of inline points generated in each iteration is counted. When the number of iterations is greater than the first iteration threshold, the plane corresponding to the iteration with the largest number of inline points is selected as the final fitting plane, that is, the plane plate extracted by segmentation, and then the inline points in the input point cloud data are removed.

Citation Information

Cited By

  • Bridge steel tower component size detection method and system based on three-dimensional laser scanning

    CN119760818A

  • Large-segment steel truss girder manufacturing configuration point cloud fusion measurement method

    CN120063120A

  • Steel truss extraction method based on point cloud segmentation and feature analysis

    CN121505371A