A data acquisition method for digital pre-assembly of engineering structures

Abstract

The present invention discloses a data acquisition method for digital pre-assembly of engineering structures, comprising the following steps: 1) placing a target ball next to a bolt hole group of a component, scanning the component using a handheld three-dimensional laser scanner, and obtaining point cloud data of the bolt hole group including the target ball; 2) using a land-based three-dimensional laser scanner to scan the component as a whole, and obtaining point cloud data of the component as a whole including the target ball; 3) using a ball feature detection algorithm to detect the target ball point cloud data; 4) performing sphere fitting on the target ball point cloud data obtained in step 3), and obtaining a set of sphere center coordinates; 5) automatically aligning the bolt hole group point cloud data obtained in step 1) and the component as a whole point cloud data obtained in step 2), and obtaining the component as a whole point cloud data and the bolt hole group point cloud data in a unified coordinate system. The present invention solves the problem that the size of the bridge steel component is large but the scanning accuracy requirement of the bolt hole group is high, and provides high-precision point cloud data for subsequent digital pre-assembly.

Description

Technical Field

[0001] The invention relates to the field of information technology and bridge safety management, and in particular to a data collection method for digital pre-assembly of engineering structures. Background Art

[0002] 3D laser scanning technology obtains geometric information of the object surface by measuring the distance between the object being measured and the scanner. The large number of data points obtained by scanning is called point cloud data. By processing the point cloud data, a 3D model of the object being measured can be obtained, which can then be used to extract object size information and other tasks. 3D laser scanners can be roughly divided into handheld and land-based 3D laser scanners based on measurement accuracy and scanning distance. The scanning range of handheld 3D laser scanners is generally within 10 meters, and the scanning accuracy can reach less than 0.1 mm; the scanning range of land-based 3D laser scanners can generally reach tens of meters or even hundreds of meters, but the average distance measurement error is usually 1 to 2 mm.

[0003] When constructing bridge projects, bridge steel components generally need to be processed in sections in the factory and transported to the site for assembly using connecting plates and high-strength bolts. However, there are often processing errors in components processed in sections. When the processing error of the bolt hole group is too large to connect the connecting plate and the component, the connecting plate needs to be rectified. Therefore, before transporting the components to the construction site, physical pre-assembly is often carried out in the factory to ensure that the components can be reliably connected. However, the cost of physical pre-assembly is high, it takes up a lot of space and has low efficiency.

[0004] The point cloud data of bridge steel components can be obtained through 3D laser scanning technology, so as to realize the digital pre-assembly of engineering structures and avoid physical pre-assembly. However, since the allowable error of the bolt holes between bridge steel components and connecting plates is only 1 mm, a high-precision handheld 3D laser scanner is required to achieve accurate dimensional measurement of the bolt hole groups of the components. However, the length of bridge steel components is generally more than 5 meters or even 10 meters. It is often impossible to scan the bolt hole groups at both ends of the component simultaneously using only a handheld 3D laser scanner. A land-based 3D laser scanner is required to coordinate the point cloud data of the bolt hole groups at both ends of the component.

[0005] Therefore, it is necessary to develop a detection method that can simultaneously scan the bolt hole groups at both ends of the bridge steel structure and meet the allowable error of the bolt holes between the bridge steel structure and the connecting plate. Summary of the invention

[0006] The purpose of the present invention is to provide a data collection method for digital pre-assembly of engineering structures to solve the problems existing in the prior art.

[0007] The technical solution adopted to achieve the purpose of the present invention is as follows: a data collection method for digital pre-assembly of engineering structures, comprising the following steps:

[0008] 1) Place at least three target balls next to the bolt hole group of the component to be assembled and cover them with scanning mark points. Use a handheld 3D laser scanner to scan the bolt hole group and target balls of the component to obtain point cloud data of the bolt hole group including the target balls.

[0009] 2) Use a land-based 3D laser scanner to scan the component as a whole to obtain the overall point cloud data of the component including the target sphere;

[0010] 3) For the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2), a ball feature detection algorithm is used to perform target ball point cloud data detection;

[0011] 4) Perform sphere fitting on the target sphere point cloud data obtained in step 3) based on the RANSAC algorithm to obtain a sphere center coordinate set;

[0012] 5) According to the spherical center coordinate set obtained in step 4), the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2) are automatically aligned to obtain the component overall point cloud data and the bolt hole group point cloud data in a unified coordinate system.

[0013] Further, step 3) comprises the following sub-steps:

[0014] 3-1) Calculate the set of neighboring points of each data point whose distance is less than the threshold t;

[0015] 3-2) Calculate the covariance matrix formed by these neighborhood points and find its three eigenvalues ​​λ1, λ2, λ3 with λ1>λ2>λ3;

[0016] 3-3) Calculate the ball feature detection operator S λ :

[0017]

[0018] Where: S λ Greater than the characteristic threshold S t It is a spherical feature point, otherwise it is a non-spherical feature point.

[0019] Further, step 4) comprises the following sub-steps:

[0020] 4-1) Randomly sample 4 data points in the target sphere point cloud data, calculate the three-dimensional sphere equation to obtain the sphere model;

[0021] 4-2) Calculate the distance from all data points to the fitted sphere, according to the distance threshold R minStatistics less than the distance threshold R min The number of data points N;

[0022] 4-3) Repeat steps 4-1) and 4-2) M times, and select the sphere model with the largest N value;

[0023] 4-4) Obtain the coordinates of the center of the sphere based on the sphere model.

[0024] Furthermore, in step 5), the two sets of point cloud data are automatically aligned in the following manner:

[0025] When there are three common and non-collinear target balls between the bolt hole group point cloud data and the component overall point cloud data, there are 6 matching methods between the two sets of point cloud data. The rotation matrix and translation matrix of each matching method are obtained by the following formula:

[0026]

[0027] Where: p i and q i are the coordinates of any sphere center in the two sets of point cloud data, R is the rotation matrix, t is the translation matrix, n is the number of sphere centers, n=3; select the registration method with the smallest matching error from the 6 matching methods;

[0028] When there are four or more common target spheres between the bolt hole group point cloud data and the component overall point cloud data, the 4PCS algorithm is used to calculate the transformation matrix:

[0029] Take a set of point cloud data as the target data, find two pairs of sphere centers and calculate the length of two line segments, find the intersection of two straight lines and the proportion of the intersection in the two line segments; search for sphere center point pairs that satisfy the length of two line segments and the proportion of the intersection in the sphere center data of another set of point cloud data; use the correspondence between the sphere centers in the two sets of point cloud data to calculate the transformation matrix, check whether the pairing is successful, and repeat multiple times to output the optimal transformation matrix.

[0030] The technical effect of the present invention is unquestionable. The method of the present invention simultaneously uses handheld and land-based three-dimensional laser scanners to scan multiple bridge steel components to be assembled, and then aligns multiple sets of point cloud data to obtain overall point cloud data of the components with detailed information of the bolt hole groups, which effectively solves the problem that the bridge steel components are large in size but the scanning accuracy requirements of the bolt hole groups are high, and provides high-precision point cloud data for subsequent digital pre-assembly. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram of collecting point cloud data of bolt hole groups using a handheld 3D laser scanner;

[0032] Figure 2This is a schematic diagram of collecting point cloud data of the entire component using a land-based 3D laser scanner;

[0033] Figure 3 The registration results of point cloud data obtained by handheld 3D laser scanner and point cloud data obtained by terrestrial 3D laser scanner;

[0034] Figure 4 Obtain point cloud data registration results for handheld 3D laser scanners;

[0035] Figure 5 This is the detailed view of the bolt hole group point cloud data;

[0036] Figure 6 is the error between the true value of the bolt hole diameter and the fitted value of the point cloud data;

[0037] Figure 7 is the distribution of bolt hole diameter error. DETAILED DESCRIPTION

[0038] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

[0039] Embodiment 1:

[0040] This embodiment discloses a data collection method for digital pre-assembly of engineering structures, comprising the following steps:

[0041] 1) Place at least three target balls next to the bolt hole group of the component to be assembled and cover them with scanning mark points. Use a handheld 3D laser scanner to scan the bolt hole group and target balls of the component to obtain point cloud data of the bolt hole group including the target balls.

[0042] When implementing it, Figure 1 As shown, the bolt hole groups of the bridge steel structure to be scanned are covered with marking points and at least three non-collinear target balls are placed, and a handheld 3D laser scanner is used to scan the bolt hole groups on both sides and the surrounding target balls.

[0043] 2) Use a land-based 3D laser scanner to scan the component as a whole to obtain the overall point cloud data of the component including the target sphere;

[0044] When implementing it, Figure 2 As shown, a land-based 3D laser scanner is used to cover and scan all the target balls on the steel components of the bridge.

[0045] 3) using a ball feature detection algorithm to perform target ball point cloud data detection on the bolt hole group point cloud data and the component overall point cloud data obtained in steps 1) and 2);

[0046] In the specific implementation, the ball feature detection algorithm is used to detect the target ball point cloud data. The basic principle of the ball feature detection algorithm is described as follows: first, the set of neighborhood points of each data point less than the distance threshold t is calculated; second, the covariance matrix composed of these neighborhood points is calculated and its three eigenvalues ​​λ1, λ2, λ3 are calculated and λ1>λ2>λ3; finally, the ball feature detection operator S is calculated. λ :

[0047]

[0048] Where S λ Greater than the characteristic threshold S t That is, it is a ball feature point, otherwise, it is a non-ball feature point. In this embodiment, the distance threshold t is 5 cm, and the feature threshold S t is 0.15.

[0049] 4) Based on the RANSAC algorithm, a sphere is fitted to the target sphere point cloud data obtained in step 3) to obtain a sphere center coordinate set;

[0050] In the specific implementation, the RANSAC algorithm is used to fit the target ball point cloud data into a sphere. The basic principle of the RANSAC algorithm is described as follows: first, four data points are randomly sampled and the three-dimensional sphere equation is calculated to obtain a sphere model; second, the distance from all data points to the fitting sphere is calculated, and the distance threshold R is used to calculate the distance between the sphere and the fitting sphere. min Statistics less than the distance threshold R min The number of data points is N; repeat M times, and select the sphere model with the largest N value; finally, obtain the coordinates of the sphere center according to the sphere model; in this embodiment, take the distance threshold R min is 5 mm, and the number of calculations M is 500 times.

[0051] 5) automatically aligning the bolt hole group point cloud data obtained in steps 1) and 2) with the component overall point cloud data according to the spherical center coordinate set obtained in step 4) to obtain the component overall point cloud data and the bolt hole group point cloud data in a unified coordinate system;

[0052] In specific implementation, if there are three common and non-collinear target balls between two sets of point cloud data, there are 6 matching methods between the two sets of point clouds. The rotation matrix and translation matrix of each matching method can be obtained by the following formula:

[0053]

[0054] Where: p i and q iare the coordinates of any sphere center in the two sets of point cloud data, R is the rotation matrix, t is the translation matrix, n is the number of sphere centers, n=3; the registration method with the smallest matching error is selected from the 6 matching methods.

[0055] If there are no less than four common target balls between the two sets of point cloud data, the 4PCS algorithm is used to calculate the transformation matrix. The basic principle of the 4PCS algorithm is described as follows: first, take a set of point cloud data as the target data, find two pairs of ball centers and calculate the lengths of the two line segments, as well as the intersection of the two straight lines and the proportion of the intersection in the two line segments; search for a pair of ball center points that satisfy the lengths of the two line segments and the proportion of the intersection in the ball center data of the other set of point cloud data; finally, use the corresponding relationship between the ball centers in the two sets of point cloud data to calculate the transformation matrix, detect whether the pairing is successful, and repeat multiple times to output the best transformation matrix;

[0056] In this embodiment, the number of common target spheres between two sets of point cloud data is three, so the above formula is used to calculate the transformation matrix; Figure 3 This is the registration result between the point cloud data obtained by the handheld 3D laser scanner and the point cloud data obtained by the terrestrial 3D laser scanner. Figure 4 The registration result between the point cloud data obtained by the handheld 3D laser scanner. Figure 5 This is a detailed view of the point cloud data of the bolt hole group.

[0057] 32 bolt holes were randomly selected and their diameters were measured with a vernier caliper. The average value of the four measurements was taken as the true value of the bolt hole diameter. The true value of the bolt hole diameter was compared with the simulated value of the point cloud data. The error is as follows: Figure 6 and 7 As shown, 90% of the errors are less than 0.2 mm, and the average error is 0.1 mm, which is less than the permissible error of 1 mm required by the engineering.

[0058] The above experimental results prove that the data acquisition method for digital pre-assembly of engineering structures disclosed in this embodiment can simultaneously obtain the overall point cloud data of bridge steel components and the local point cloud data of bolt hole groups, and meet the requirements of engineering accuracy. Therefore, the method described in this embodiment is practical and effective.

[0059] Embodiment 2:

[0060] This embodiment discloses a data collection method for digital pre-assembly of engineering structures, comprising the following steps:

[0061] 1) Place at least three target balls next to the bolt hole group of the component to be assembled and cover them with scanning mark points. Use a handheld 3D laser scanner to scan the bolt hole group and target balls of the component to obtain the point cloud data of the bolt hole group including the target balls; see Figure 1, is a schematic diagram of scanning the bolt hole group and target ball of a component using a handheld 3D laser scanner;

[0062] 2) Use a land-based 3D laser scanner to scan the component as a whole to obtain the overall point cloud data of the component including the target sphere; see Figure 2 , which is a schematic diagram of using a land-based 3D laser scanner to scan the entire component.

[0063] 3) For the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2), a ball feature detection algorithm is used to perform target ball point cloud data detection;

[0064] 4) Perform sphere fitting on the target sphere point cloud data obtained in step 3) based on the RANSAC algorithm to obtain a sphere center coordinate set;

[0065] 5) According to the spherical center coordinate set obtained in step 4), the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2) are automatically aligned to obtain the component overall point cloud data and the bolt hole group point cloud data in a unified coordinate system. Figure 3 This is the registration result between the point cloud data obtained by the handheld 3D laser scanner and the point cloud data obtained by the terrestrial 3D laser scanner. Figure 4 The registration result between the point cloud data obtained by the handheld 3D laser scanner. Figure 5 This is a detailed view of the point cloud data of the bolt hole group.

[0066] Embodiment 3:

[0067] The main steps of this embodiment are the same as those of Embodiment 2. Further, step 3) includes the following sub-steps:

[0068] 3-1) Calculate the set of neighboring points of each data point whose distance is less than the threshold t;

[0069] 3-2) Calculate the covariance matrix formed by these neighborhood points and find its three eigenvalues ​​λ1, λ2, λ3 with λ1>λ2>λ3;

[0070] 3-3) Calculate the ball feature detection operator S λ :

[0071]

[0072] Where: S λ Greater than the characteristic threshold S t It is a spherical feature point, otherwise it is a non-spherical feature point.

[0073] Embodiment 4:

[0074] The main steps of this embodiment are the same as those of Embodiment 2. Further, step 4) includes the following sub-steps:

[0075] 4-1) Randomly sample 4 data points in the target sphere point cloud data, calculate the three-dimensional sphere equation to obtain the sphere model;

[0076] 4-2) Calculate the distance from all data points to the fitted sphere, according to the distance threshold R min Statistics less than the distance threshold R min The number of data points N;

[0077] 4-3) Repeat steps 4-1) and 4-2) M times, and select the sphere model with the largest N value;

[0078] 4-4) Obtain the coordinates of the center of the sphere based on the sphere model.

[0079] Embodiment 5:

[0080] The main steps of this embodiment are the same as those of Embodiment 2. Furthermore, in step 5), the following method is used to automatically align the two sets of point cloud data:

[0081] When there are three common and non-collinear target balls between the bolt hole group point cloud data and the component overall point cloud data, there are 6 matching methods between the two sets of point cloud data. The rotation matrix and translation matrix of each matching method are obtained by the following formula:

[0082]

[0083] Where: p i and q i are the coordinates of any sphere center in the two sets of point cloud data, R is the rotation matrix, t is the translation matrix, n is the number of sphere centers, n=3; the registration method with the smallest matching error is selected from the 6 matching methods.

[0084] When there are four or more common target spheres between the bolt hole group point cloud data and the component overall point cloud data, the 4PCS algorithm is used to calculate the transformation matrix:

[0085] Take a set of point cloud data as the target data, find two pairs of sphere centers and calculate the length of two line segments, find the intersection of two straight lines and the proportion of the intersection in the two line segments; search for sphere center point pairs that satisfy the length of two line segments and the proportion of the intersection in the sphere center data of another set of point cloud data; use the correspondence between the sphere centers in the two sets of point cloud data to calculate the transformation matrix, check whether the pairing is successful, and repeat multiple times to output the optimal transformation matrix.

Claims

1. A data collection method for digital pre-assembly of engineering structures, characterized in that: The following steps are involved: 1) Place at least three target balls next to the bolt hole group of the component to be assembled and cover them with scanning mark points. Use a handheld 3D laser scanner to scan the bolt hole group and target balls of the component to obtain point cloud data of the bolt hole group including the target balls. 2) Use a land-based 3D laser scanner to scan the component as a whole to obtain the overall point cloud data of the component including the target sphere; 3) For the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2), a ball feature detection algorithm is used to perform target ball point cloud data detection; step 3) includes the following sub-steps: 3-1) Calculate the set of neighboring points of each data point whose distance is less than the threshold t; 3-2) Calculate the covariance matrix formed by these neighborhood points and find its three eigenvalues ​​λ1, λ2, λ3 with λ1>λ2>λ3; 3-3) Calculate the ball feature detection operator S λ : Where: S λ Greater than the characteristic threshold S t It is a spherical feature point, otherwise, it is a non-spherical feature point; 4) Perform sphere fitting on the target sphere point cloud data obtained in step 3) based on the RANSAC algorithm to obtain a sphere center coordinate set; step 4) includes the following sub-steps: 4-1) Randomly sample 4 data points in the target sphere point cloud data, calculate the three-dimensional sphere equation to obtain the sphere model; 4-2) Calculate the distance from all data points to the fitted sphere, according to the distance threshold R min Statistics less than the distance threshold R min The number of data points N; 4-3) Repeat steps 4-1) and 4-2) M times, and select the sphere model with the largest N value; 4-4) Obtaining the coordinates of the sphere center according to the sphere model; 5) According to the spherical center coordinate set obtained in step 4), the bolt hole group point cloud data obtained in step 1) and the component overall point cloud data obtained in step 2) are automatically aligned to obtain the component overall point cloud data and the bolt hole group point cloud data in a unified coordinate system; in step 5), the two sets of point cloud data are automatically aligned in the following manner: When there are three common and non-collinear target balls between the bolt hole group point cloud data and the component overall point cloud data, there are 6 matching methods between the two sets of point cloud data. The rotation matrix and translation matrix of each matching method are obtained by the following formula: Where: p i and q i are the coordinates of any sphere center in the two sets of point cloud data, R is the rotation matrix, t is the translation matrix, n is the number of sphere centers, n=3; select the registration method with the smallest matching error from the 6 matching methods; When there are four or more common target spheres between the bolt hole group point cloud data and the component overall point cloud data, the 4PCS algorithm is used to calculate the transformation matrix: Take a set of point cloud data as the target data, find two pairs of sphere centers and calculate the length of two line segments, find the intersection of two straight lines and the proportion of the intersection in the two line segments; search for sphere center point pairs that satisfy the length of two line segments and the proportion of the intersection in the sphere center data of another set of point cloud data; use the correspondence between the sphere centers in the two sets of point cloud data to calculate the transformation matrix, check whether the pairing is successful, and repeat multiple times to output the optimal transformation matrix.

Images