A component deviation measurement method, device, system, and storage medium

Abstract

The application provides a kind of component deviation measurement method, device, system and storage medium, belong to component detection technical field, method includes: by three-dimensional laser scanner to the main arch rib steel pipe component to be measured is scanned and obtains original point cloud data;Original point cloud data is filtered to obtain filtered point cloud data;Filtering after point cloud data is analyzed to obtain view point feature histogram feature descriptor and target feature point;According to the view point feature histogram feature descriptor, the target feature point is registered and analyzed to obtain target point cloud data;Through component reference model, the deviation measurement of target point cloud data is obtained to obtain component deviation measurement result.The application improves the stability and accuracy of registration, effectively avoids the risk of potential dangerous construction area operation, ensures the normal operation of construction production activities, avoids any interference, and ensures the smooth progress of construction progress.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of component detection, and in particular to a component deviation measurement method, device, system and storage medium. Background Art

[0002] In steel-concrete-filled arch bridges, the main arch rib steel tubes are the primary load-bearing components. Their primary characteristics are their circular shape, thick walls, and large area and size. They are often used in large spans, such as across rivers and in mountainous canyons. Their spans range from 100 to 500 meters, and their weights range from tens to thousands of tons. To prevent dimensional errors in the main arch rib steel tubes from affecting project quality, traditional manual measurement methods are often used. Total stations are used at the prefabrication site to measure relevant dimensional points and determine component deviations to ensure the main arch rib steel tubes meet design requirements. However, this method is inefficient and can pose safety risks, contrary to the principles of safe production operations.

[0003] Compared to traditional measurement methods, component deviation measurement using point cloud data generated by 3D laser scanners has become widely used in industry. However, point cloud registration is essential for obtaining complete component point cloud data during point cloud data processing. The most classic point cloud registration algorithm is the ICP algorithm, but this algorithm places high demands on the initial measurement position of the 3D laser scanner. Otherwise, it easily falls into a local optimal solution, resulting in matching failures. Ultimately, component deviation measurement accuracy is low, reaching only 2-3 cm, which clearly does not meet the measurement requirements for the main arch rib steel pipes of steel-concrete-filled arch bridges. Main arch rib steel pipes are measured at the precast site, subject to interference from numerous equipment and other steel pipes. Finding an ideal initial measurement position is often difficult, necessitating the establishment of multiple measurement stations and the splicing of point cloud data from multiple stations. Therefore, relying on existing algorithms and technologies, 3D laser measurement of the main arch rib steel pipes of steel-concrete-filled arch bridges at the precast site does not meet the project's accuracy requirements.

[0004] In view of this, it is urgent to develop a component deviation measurement method to improve the stability and accuracy of alignment. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a component deviation measurement method, device, system and storage medium in response to the deficiencies of the existing technology.

[0006] The present invention solves the above technical problems with the following technical solutions: A component deviation measurement method comprises the following steps:

[0007] The main arch rib steel pipe component to be measured is scanned by a 3D laser scanner to obtain multiple original point cloud data;

[0008] Performing filtering on all the original point cloud data to obtain a plurality of filtered point cloud data;

[0009] Performing feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points;

[0010] Performing registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data;

[0011] A component reference model is imported, and deviation measurement is performed on all target point cloud data using the component reference model to obtain component deviation measurement results.

[0012] Another technical solution of the present invention to solve the above technical problem is as follows: a component deviation measuring device, comprising:

[0013] A scanning module is used to scan the main arch rib steel pipe component to be measured using a 3D laser scanner to obtain multiple original point cloud data;

[0014] A filtering module, configured to filter all the original point cloud data to obtain a plurality of filtered point cloud data;

[0015] A feature analysis module is used to perform feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points;

[0016] A registration analysis module, configured to perform registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data;

[0017] The measurement result acquisition module is used to import a component reference model, perform deviation measurement on all the target point cloud data through the component reference model, and obtain component deviation measurement results.

[0018] Based on the above-mentioned component deviation measurement method, the present invention also provides a component deviation measurement system.

[0019] Another technical solution of the present invention to solve the above technical problems is as follows: a component deviation measurement system includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the component deviation measurement method described above is implemented.

[0020] Based on the above-mentioned component deviation measurement method, the present invention also provides a computer-readable storage medium.

[0021] Another technical solution of the present invention to solve the above technical problem is as follows: a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the component deviation measurement method as described above is implemented.

[0022] The beneficial effects of the present invention are as follows: original point cloud data is obtained by scanning the main arch rib steel pipe component to be measured with a three-dimensional laser scanner, filtered point cloud data is obtained by filtering the original point cloud data, feature analysis of the filtered point cloud data is obtained to obtain a viewpoint feature histogram feature descriptor and target feature points, target point cloud data is obtained by alignment analysis of the target feature points according to the viewpoint feature histogram feature descriptor, component deviation measurement results are obtained by measuring the deviation of the target point cloud data with a component reference model, the stability and accuracy of the alignment are improved, the risk of operations in potentially dangerous construction areas is effectively avoided, the normal progress of construction production activities is ensured, any interference is avoided, the smooth progress of construction progress is guaranteed, the efficiency of data acquisition, data accuracy and the safety of measurement operations are improved, and at the same time, the operating personnel are enabled to perform dimensional measurements outside the dangerous area, while ensuring the measurement accuracy of the dimensions and the calculation of the deviation of the main arch rib steel pipe component, the safety of the operating personnel is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic flow chart of a component deviation measurement method provided in an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of deviation analysis of a component deviation measurement method provided in an embodiment of the present invention;

[0025] Figure 3 This is a module block diagram of a component deviation measurement device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0027] Figure 1 A schematic flow chart of a component deviation measurement method provided in an embodiment of the present invention.

[0028] like Figure 1 As shown, a component deviation measurement method includes the following steps:

[0029] The main arch rib steel pipe component to be measured is scanned by a 3D laser scanner to obtain multiple original point cloud data;

[0030] Performing filtering on all the original point cloud data to obtain a plurality of filtered point cloud data;

[0031] Performing feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points;

[0032] Performing registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data;

[0033] A component reference model is imported, and deviation measurement is performed on all target point cloud data using the component reference model to obtain component deviation measurement results.

[0034] It should be understood that the scanning point cloud data P and Q (ie, original point cloud data) of the main arch rib steel pipe component to be measured are obtained by using the front and rear stations of the three-dimensional laser scanner respectively.

[0035] It should be understood that the scanned point cloud component (ie, original point cloud data) is pre-processed by the nearest point voxel filtering algorithm to obtain point cloud data p, q (ie, filtered point cloud data).

[0036] Specifically, the registered point cloud data of the main arch rib steel tube component (i.e., the target point cloud data) and the CAD design reference model (i.e., the component reference model) were loaded into the Geomagic software together. The point cloud scanning data of the main arch rib steel tube component (i.e., the target point cloud data) and the CAD design reference model (i.e., the component reference model) were analyzed and compared to obtain an analysis report on the external dimension deviation (i.e., the component deviation measurement results).

[0037] Specifically, the specific process of using a 3D laser scanner to obtain point cloud data is as follows:

[0038] Clean the surface of the main arch rib steel pipe component to avoid any obstruction between the main arch rib steel pipe component and the 3D laser scanner; because the material of the main arch rib steel pipe component is affected by the temperature environment, the 3D laser scanner chooses to operate at night and early morning to ensure that the environment is 20°C (the allowable temperature difference is within ±2°); determine the relationship between several scanning stations and their positions based on the shape and height of the main arch rib steel pipe component to ensure that the scanning range of the 3D laser scanner at several scanning stations can cover the main arch rib steel pipe component; adjust the corresponding parameters of the 3D laser scanner according to the operating environment of the 3D laser scanner; based on the parameter adjustment, the 3D laser scanner scans the main arch rib steel pipe component at each scanning station to obtain the point cloud scanning data P and Q (i.e., original point cloud data) of the main arch rib steel pipe component.

[0039] It should be understood that the component reference model can be obtained through SolidWorks software.

[0040] In the above embodiment, the original point cloud data is obtained by scanning the main arch rib steel pipe component to be measured by a three-dimensional laser scanner, the filtered point cloud data is obtained by filtering the original point cloud data, the view point feature histogram feature descriptor and the target feature point are obtained by analyzing the features of the filtered point cloud data, the target point cloud data is obtained by registering and analyzing the target feature point according to the view point feature histogram feature descriptor, the component deviation measurement result is obtained by measuring the deviation of the target point cloud data by the component reference model, the stability and accuracy of registration are improved, the risk of operation in a potential dangerous construction area is effectively avoided, the normal operation of construction and production activities is ensured, any interference is avoided, the smooth progress of construction progress is ensured, the data acquisition efficiency, data accuracy and measurement operation safety are improved, at the same time, the size measurement of the operating personnel in the dangerous area is realized, the safety of the operating personnel is ensured while ensuring the measurement size accuracy and calculating the deviation of the main arch rib steel pipe component.

[0041] Optionally, as an embodiment of the present application, the process of analyzing the features of all the filtered point cloud data to obtain the view point feature histogram feature descriptor and the plurality of target feature points comprises:

[0042] Analyzing the features of all the filtered point cloud data to obtain a plurality of target feature points;

[0043] Analyzing the features of all the target feature points to obtain the view point feature histogram feature descriptor.

[0044] In the above embodiment, the view point feature histogram feature descriptor and the plurality of target feature points are obtained by analyzing the features of all the filtered point cloud data, the stability and accuracy of registration are improved, the risk of operation in a potential dangerous construction area is effectively avoided, the normal operation of construction and production activities is ensured, any interference is avoided, the smooth progress of construction progress is ensured, the data acquisition efficiency, data accuracy and measurement operation safety are improved.

[0045] Optionally, as an embodiment of the present application, the process of analyzing the features of all the filtered point cloud data to obtain the plurality of target feature points comprises:

[0046] Analyzing the local features of all the filtered point cloud data to obtain a plurality of first initial feature points;

[0047] Analyzing the weighted vector sine values of all the filtered point cloud data, taking the analysis result as a second initial feature point set, and taking the first initial feature points and the feature points in the second initial feature point set as target feature points, thereby obtaining a plurality of target feature points.

[0048] It should be understood that the feature points of the point cloud of the main arch rib steel pipe component (ie, the first initial feature points) are extracted.

[0049] Specifically, for the area where the surface of the main arch rib steel tube component changes gently, the point (i.e., the second initial feature point) calculated using the sine value of the weighted vector at different scales is combined with the obtained feature point (i.e., the first initial feature point) to form the final feature point (i.e., the target feature point).

[0050] In the above embodiment, feature point analysis is performed on all the filtered point cloud data to obtain multiple target feature points, which improves the stability and accuracy of the alignment, effectively avoids the risks of operations in potentially dangerous construction areas, ensures the normal progress of construction and production activities, avoids any interference, ensures the smooth progress of construction, and improves data acquisition efficiency, data accuracy and the safety of measurement operations.

[0051] Optionally, as an embodiment of the present invention, the process of performing local feature analysis on all the filtered point cloud data to obtain a plurality of first initial feature points includes:

[0052] Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset first radius to obtain a plurality of first neighborhood points corresponding to each of the filtered point cloud data;

[0053] The first point cloud weights are calculated for each of the filtered point cloud data and the multiple first neighborhood points corresponding to each of the filtered point cloud data using the first formula to obtain multiple first point cloud weights corresponding to each of the filtered point cloud data. The first formula is:

[0054]

[0055] Among them, w1 ij is the first point cloud weight corresponding to the jth first neighborhood point of the i-th filtered point cloud data, p i is the i-th filtered point cloud data, is the jth first neighborhood point corresponding to the i-th filtered point cloud data;

[0056] The covariance matrix corresponding to each of the filtered point cloud data is calculated by the second formula for each of the preset first radius, each of the filtered point cloud data, a plurality of first neighborhood points corresponding to each of the filtered point cloud data, and a plurality of first point cloud weights corresponding to each of the filtered point cloud data, to obtain a covariance matrix corresponding to each of the filtered point cloud data. The second formula is:

[0057]

[0058] Among them, C iis the covariance matrix corresponding to the i-th filtered point cloud data, r iss The preset first radius, w1 ij is the first point cloud weight corresponding to the jth first neighborhood point of the i-th filtered point cloud data, p i is the i-th filtered point cloud data, is the first neighborhood point corresponding to the i-th filtered point cloud data;

[0059] Performing eigenvalue decomposition on each of the covariance matrices to obtain a first initial eigenvalue corresponding to each filtered point cloud data, a second initial eigenvalue corresponding to each filtered point cloud data, and a third initial eigenvalue corresponding to each filtered point cloud data;

[0060] sorting the first initial eigenvalues, the second initial eigenvalues ​​corresponding to the filtered point cloud data, and the third initial eigenvalues ​​corresponding to the filtered point cloud data in ascending order to obtain first sorted eigenvalues ​​corresponding to the filtered point cloud data, second sorted eigenvalues ​​corresponding to the filtered point cloud data, and third sorted eigenvalues ​​corresponding to the filtered point cloud data;

[0061] When the first sorted eigenvalue, the second sorted eigenvalue, and the third sorted eigenvalue satisfy the third formula and the fourth formula, the filtered point cloud data corresponding to the first sorted eigenvalue is used as the first initial feature point, thereby obtaining multiple first initial feature points. The third formula is:

[0062]

[0063] Among them, λ i2 is the second sorted eigenvalue corresponding to the i-th filtered point cloud data, λ i3 is the third sorted eigenvalue corresponding to the i-th filtered point cloud data, and ε1 is the preset first characteristic threshold;

[0064] The fourth formula is:

[0065]

[0066] Among them, λ i1 is the first sorted eigenvalue corresponding to the i-th filtered point cloud data, λ i2 is the second sorted eigenvalue corresponding to the i-th filtered point cloud data, and ε2 is the preset second feature threshold.

[0067] It should be understood that for a point pi in a given point cloud data P (i.e., filtered point cloud data), the KD-tree algorithm is used to perform neighborhood search to obtain the radius r of point pi. iss The point set p in the neighborhood spacej (i.e. multiple first neighbor points).

[0068] It should be understood that the K-dimensional tree algorithm (KD-Tree algorithm) is a data structure for organizing points in K-dimensional space, and is particularly suitable for efficient search of multidimensional data. The KD tree recursively divides the hyperplane in the K-dimensional space into two parts, forming a binary tree structure, and each node represents a point in the K-dimensional space. This data structure can significantly improve the efficiency of searching in multidimensional space, and is particularly suitable for operations such as range search and nearest neighbor search. By constructing a balanced KD tree, a specific point or a group of points in the space can be quickly located, which is widely used in many fields such as computer vision, robot navigation, and database retrieval.

[0069] Specifically, the Euclidean distance between the query point pi (i.e., the filtered point cloud data) and each point in the neighborhood (i.e., multiple first neighborhood points) is calculated, and the weight w is calculated. ij (i.e. the first point cloud weight), the specific formula is as follows:

[0070]

[0071] Calculate the covariance matrix C between the query point pi (i.e., filtered point cloud data) and all points in the neighborhood (i.e., multiple first neighborhood points). The specific formula is as follows:

[0072]

[0073] Perform eigenvalue decomposition on the covariance matrix C to obtain {λ1, λ2, λ3}, satisfying λ1<λ2<λ3.

[0074] Preferably, the threshold values ​​ε1 = 0.985 and ε2 = 0.985 are set.

[0075] Specifically, a threshold ε1 (i.e., a preset first feature threshold) and ε2 (i.e., a preset second feature threshold) are set, and the point pi is considered to be a feature point (i.e., the first initial feature point) if the following formula is satisfied. The specific formula is as follows:

[0076]

[0077] In the above embodiment, local feature analysis is performed on all filtered point cloud data to obtain multiple first initial feature points, which improves the stability and accuracy of the alignment, effectively avoids the risks of operations in potentially dangerous construction areas, ensures the normal progress of construction and production activities, avoids any interference, ensures the smooth progress of construction, and improves data acquisition efficiency, data accuracy and the safety of measurement operations.

[0078] Optionally, as an embodiment of the present invention, the process of analyzing the weighted vector sine value of all the filtered point cloud data and using the analysis result as the second initial feature point set includes:

[0079] Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset second radius to obtain a plurality of second neighborhood points corresponding to each of the filtered point cloud data;

[0080] Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset third radius to obtain a plurality of third neighborhood points corresponding to each of the filtered point cloud data;

[0081] Vector calculations are performed on each of the filtered point cloud data, the multiple second neighborhood points corresponding to each of the filtered point cloud data, and the multiple third neighborhood points corresponding to each of the filtered point cloud data using a first set of equations to obtain multiple first neighborhood point vectors corresponding to each of the filtered point cloud data and multiple second neighborhood point vectors corresponding to each of the filtered point cloud data. The first set of equations is:

[0082]

[0083] in, is the first neighborhood point vector corresponding to the ath second neighborhood point of the i-th filtered point cloud data, pi is the i-th filtered point cloud data, is the ath second neighborhood point corresponding to the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data, is the bth third neighborhood point corresponding to the i-th filtered point cloud data;

[0084] The second point cloud weights are calculated for each of the plurality of first neighborhood point vectors corresponding to each of the filtered point cloud data by using the fifth formula to obtain a plurality of second point cloud weights corresponding to each of the filtered point cloud data. The fifth formula is:

[0085]

[0086] Among them, w2 ia is the second point cloud weight corresponding to the a-th second neighborhood point of the i-th filtered point cloud data, is the first neighborhood point vector corresponding to the a-th second neighborhood point of the i-th filtered point cloud data;

[0087] The third point cloud weights are calculated for each of the plurality of second neighborhood point vectors corresponding to each of the filtered point cloud data using the sixth formula to obtain a plurality of third point cloud weights corresponding to each of the filtered point cloud data. The sixth formula is:

[0088]

[0089] Among them, w3 ib is the third point cloud weight corresponding to the bth third neighborhood point of the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data;

[0090] The neighborhood angles are calculated for each of the first neighborhood point vectors corresponding to each of the filtered point cloud data, the second neighborhood point vectors corresponding to each of the filtered point cloud data, the second point cloud weights corresponding to each of the filtered point cloud data, and the third point cloud weights corresponding to each of the filtered point cloud data using the seventh formula to obtain an initial neighborhood angle corresponding to each of the filtered point cloud data. The seventh formula is:

[0091]

[0092] Among them, θ i is the initial neighborhood angle corresponding to the i-th filtered point cloud data, is the first neighborhood point vector corresponding to the ath second neighborhood point of the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data, w2 ia is the second point cloud weight corresponding to the ath second neighborhood point of the i-th filtered point cloud data, w3 ib is the third point cloud weight corresponding to the bth third neighborhood point of the i-th filtered point cloud data;

[0093] Sorting all the initial neighborhood angles in descending order to obtain sorted neighborhood angles corresponding to the filtered point cloud data;

[0094] Filtering out the sorted neighborhood angles that meet a screening condition from all the sorted neighborhood angles, and obtaining a plurality of filtered neighborhood angles after screening, wherein the screening condition is that the sorted neighborhood angle is less than a preset angle threshold;

[0095] The filtered point cloud data corresponding to the first A filtered neighborhood angles are used as second initial feature points, thereby obtaining multiple second initial feature points, and all of the second initial feature points are collected to obtain a second initial feature point set.

[0096] Preferably, the preset second radius may be 0.05, the preset third radius may be 0.1, and the preset angle threshold may be 5 degrees.

[0097] It should be understood that different radius neighborhoods r1 (i.e., preset second radius) and r2 (i.e., preset third radius) are set for each point in p (i.e., point cloud data after filtering); for a point pi in p (i.e., point cloud data after filtering), the point set p1x (i.e., multiple second neighborhood points) in the neighborhood space of radius r1 of point pi and the point set p2y (i.e., multiple third neighborhood points) in the neighborhood space of radius r2 are obtained through neighborhood search.

[0098] Specifically, we obtain the center point pi to the point set p 1i 、p 2i The vector of (i.e., the first neighborhood point vector and the second neighborhood point vector) is as follows:

[0099]

[0100] Calculate the sum of the vectors of all points in different neighborhoods respectively, and calculate the weight function (i.e. the weight of the second point cloud and the weight of the third point cloud). The specific formula is as follows:

[0101]

[0102] The cosine of the angle between weighted vectors is calculated using the dot product of the vectors. The specific formula is as follows:

[0103]

[0104] In the above formula, θ is the angle between the weighted points in different neighborhoods and the center point pi (i.e., the initial neighborhood angle);

[0105] Using the sine-cosine conversion relationship, the sine of the angle can be obtained using the cosine of the angle. The specific formula is as follows:

[0106]

[0107] According to the above steps, the KD-Tree algorithm is used to traverse and calculate the weighted vector sine value of each point p, and arrange them from large to small, set their angle threshold θ (i.e., the preset angle threshold) and the percentage threshold ε0, and select the points that are less than the angle threshold θ and ranked high as feature points (i.e., the second initial feature points).

[0108] It should be understood that taking the filtered point cloud data corresponding to the top A filtered neighborhood angles as the second initial feature point means taking the filtered point cloud data corresponding to the top A filtered neighborhood angles as the second initial feature point.

[0109] It should be understood that the angle threshold θ is set to 5° and the percentage threshold ε0 is set to 30%, and the points that are smaller than the angle threshold θ and ranked high are selected as feature points (ie, the second initial feature points).

[0110] In the above embodiment, the weighted vector sine value analysis is performed on all the filtered point cloud data, and the analysis result is taken as the second initial feature point set, which ensures the normal progress of the construction production activities, avoids any interference, guarantees the smooth progress of the construction progress, improves the data acquisition efficiency, data accuracy and measurement operation safety, at the same time, realizes the size measurement of the operating personnel outside the dangerous area, guarantees the safety of the operating personnel while ensuring the measurement size accuracy and calculating the deviation of the main arch rib steel pipe component.

[0111] Optionally, as an embodiment of the present application, the process of analyzing the feature descriptor of all the target feature points to obtain the view point feature histogram feature descriptor comprises:

[0112] The neighborhood search is performed on each target feature point by using the k nearest neighbor algorithm to obtain a plurality of fourth neighborhood points corresponding to each target feature point;

[0113] The normal vector calculation is performed on the plurality of fourth neighborhood points corresponding to each target feature point by using the principal component analysis algorithm to obtain a normal vector corresponding to each target feature point;

[0114] The center point calculation is performed on all the target feature points by using the seventh formula to obtain a center point The seventh formula is:

[0115]

[0116] wherein, is the X-axis coordinate of the center point, is the Y-axis coordinate of the center point, is the Z-axis coordinate of the center point, and N is the total number of target feature points, x k is the X-axis coordinate of the kth target feature point, y k is the Y-axis coordinate of the kth target feature point, z k is the Z-axis coordinate of the kth target feature point;

[0117] The center point included angle calculation is performed on the center point, each target feature point and the normal vector corresponding to each target feature point by using the eighth formula to obtain a center point included angle corresponding to each target feature point, and the eighth formula is:

[0118]

[0119] wherein, α k is the center point included angle corresponding to the kth target feature point, o is the center point, p4 k is the kth target feature point, n k is the normal vector corresponding to the kth target feature point;

[0120] The normal vector angles corresponding to the normal vectors of each target feature point and the normal vectors corresponding to any adjacent target feature point are calculated by the ninth formula to obtain multiple normal vector angles corresponding to each target feature point. The ninth formula is:

[0121]

[0122] Among them, β ku is the angle between the normal vectors of the kth target feature point and the uth target feature point, n k is the normal vector corresponding to the kth target feature point, n u is the normal vector corresponding to the u-th target feature point, arccos() is the inverse cosine function;

[0123] Performing a viewpoint feature histogram analysis on all the center point angles and all the normal vector angles to obtain a viewpoint feature histogram;

[0124] Vectorization is performed on the viewpoint feature histogram to obtain a viewpoint feature histogram feature descriptor.

[0125] It should be understood that, according to the final feature point (ie, the target feature point), the viewpoint feature histogram feature descriptor is obtained by utilizing the relationship between the normal vector, the point, and the centroid.

[0126] It should be understood that the viewpoint feature histogram feature descriptor, or VFH (Viewpoint Feature Histogram) feature descriptor, is a descriptor for point cloud data, primarily used for point cloud clustering and six-degree-of-freedom pose estimation. The VFH feature descriptor, developed based on the FPFH (Fast Point Feature Histograms) descriptor, calculates the normal vector distribution for each point in a point cloud to generate a global feature vector that describes the shape and geometric properties of the entire point cloud. VFH is rotationally invariant and robust to noise, making it widely used in tasks such as 3D object recognition, classification, and clustering.

[0127] Specifically, the k-Nearest Neighbors (KNN) algorithm is a distance-based classification and regression method. Its core idea is to find the K closest training instances to a new instance in a labeled dataset based on a distance metric. The label of the new instance is then predicted based on the information from these K neighbors. In classification problems, the most common approach is to use majority voting, where the new instance is assigned the category that appears most frequently among the K nearest neighbors. In regression problems, the prediction is typically based on the average of the target values ​​of the K nearest neighbors. The KNN algorithm can be used for both classification and regression problems.

[0128] Principal Component Analysis (PCA) is a statistical dimensionality reduction technique that uses an orthogonal transformation to transform a set of potentially correlated variables into a set of linearly uncorrelated variables. These new variables are called principal components. The goal of PCA is to reduce the dimensionality of the data and simplify the model while preserving as much information as possible from the original data.

[0129] Specifically, assuming that there are N feature points (i.e., target feature points), for each feature point indexed (i.e., target feature point), the kNN algorithm is used to find its k nearest neighboring points based on the K-dimensional tree principle. These points constitute a local neighborhood (i.e., multiple fourth neighborhood points), and the normal vector n of each feature point is calculated using the PCA algorithm. i (n xi , n yi , n zi ).

[0130] Specifically, the x, y, and z coordinates of each point are summed up to find the centroid of the entire point cloud. (i.e. the center point), the specific formula is as follows:

[0131]

[0132] Calculate the normal vector n for each point i The angles between them and the centroid (i.e., the center point angle) are respectively set to a feature point p (i.e., the target feature point). The specific formula is as follows:

[0133]

[0134] Calculate the angle between adjacent normal vectors (i.e., the normal vector angle). Let the normal vector of another feature point be v. The specific formula is as follows:

[0135]

[0136] According to the above-mentioned included angle α, φ, the VFH feature descriptor histogram (i.e. the view feature histogram feature descriptor) is generated.

[0137] It should be understood that the vector form of the histogram H (i.e. the view feature histogram) is the view feature histogram feature descriptor.

[0138] In the above-mentioned embodiment, the analysis of the feature descriptors of all target feature points obtains the view feature histogram feature descriptor, which realizes the size measurement of the work personnel outside the dangerous area, ensures the safety of the work personnel while ensuring the measurement size precision and calculating the deviation of the main arch rib steel pipe component.

[0139] Optionally, as an embodiment of the present application, the process of analyzing the view feature histogram of all the center point included angles and all the normal vector included angles to obtain the view feature histogram includes:

[0140] The minimum value is selected from all the center point included angles to obtain the minimum center point included angle; the maximum value is selected from all the center point included angles to obtain the maximum center point included angle; the minimum value is selected from all the normal vector included angles to obtain the minimum normal vector included angle; the maximum value is selected from all the normal vector included angles to obtain the maximum normal vector included angle;

[0141] The minimum center point included angle is taken as the minimum value and the maximum center point included angle is taken as the maximum value to obtain the center point included angle interval;

[0142] The minimum normal vector included angle is taken as the minimum value and the maximum normal vector included angle is taken as the maximum value to obtain the normal vector included angle interval;

[0143] The center point included angle interval is equally divided into a plurality of center point included angle sub-intervals;

[0144] The total number of all the center point included angle sub-intervals is counted to obtain the center point included angle sub-interval total number; the normal vector included angle interval is equally divided into a plurality of normal vector included angle sub-intervals;

[0145] The total number of all the normal vector included angle sub-intervals is counted to obtain the normal vector included angle sub-interval total number;

[0146] The minimum center point included angle, the maximum center point included angle, the center point included angle sub-interval total number and the plurality of center point included angles in each center point included angle sub-interval are subjected to frequency calculation through the tenth formula to obtain the center point included angle frequency corresponding to each center point included angle sub-interval, and the tenth formula is:

[0147]

[0148] wherein,

[0149] Among them, H α (j′) is the frequency of the center point angle corresponding to the j′th center point angle subinterval, N is the total number of target feature points, is the indicator function, α min is the minimum center point angle, α max is the maximum center point angle, Δα is the center point angle interval width, α k is the center point angle corresponding to the k-th target feature point, m′ is the total number of center point angle subintervals, and j′ is the j′-th center point angle subinterval;

[0150] The frequency of the normal vector angle corresponding to each normal vector angle subinterval is obtained by calculating the minimum normal vector angle, the maximum normal vector angle, the total number of normal vector angle subintervals, and the number of normal vector angles in each normal vector angle subinterval using the eleventh formula. The eleventh formula is:

[0151]

[0152] in,

[0153] Among them, H β (k′) is the frequency of the normal vector angle corresponding to the k′th normal vector angle subinterval, N is the total number of target feature points, is the indicator function, adj(k) is the set of target feature points adjacent to the kth target feature point, β min is the minimum normal vector angle, Δβ is the normal vector angle interval width, β ku is the angle between the normal vectors of the kth target feature point and the uth target feature point, k′ is the k′th normal vector angle subinterval, β max is the maximum normal vector angle, n′ is the total number of normal vector angle subintervals;

[0154] The viewpoint feature histogram is calculated for all the center point angle frequencies and all the normal vector angle frequencies using the twelfth formula to obtain the viewpoint feature histogram. The twelfth formula is:

[0155] H=|H α (1), H α (2), ..., H α (j′), H β (1), H β (2), ..., H β (k′)|,

[0156] Among them, H is the viewpoint feature histogram, H α (j′) is the frequency of the central point angle corresponding to the j′th central point angle subinterval, Hβ (k′) is the normal vector angle frequency corresponding to the k′th normal vector angle subinterval.

[0157] It should be understood that the first step is to determine the histogram dimensions for constructing the VFH feature descriptor and the method for dividing the angles into intervals. Assuming that the final VFH histogram is divided into M dimensions, the value ranges of angles α and β are divided into m and n intervals (m + n = M) with equal spacing (i.e., the center point angle subinterval and the normal vector angle subinterval).

[0158] Specifically, for α (i.e., the center point angle), let its value range be [α max , α min ]; for β (i.e. normal vector angle), let its value range be [β max , β min ] Among them, for the angle α (i.e. the center point angle), the width of each interval is Δα (i.e. the center point angle interval width), and for the angle β (i.e. the normal vector angle), the width of each interval is Δβ (i.e. the normal vector angle interval width). The specific formula is as follows:

[0159]

[0160] Then, the frequency of the angle falling into each interval is counted, which is specifically divided into angle statistics for the angle α with the center of mass line and angle statistics for the angle β between adjacent normal vectors.

[0161] Specifically, the angle statistics method for the angle α with the centroid (i.e., the frequency of the center point angle) is as follows: traverse all feature points p in the point cloud i (i=1, 2, ..., N, N is the total number of feature points) (i.e., target feature points), for each feature point, calculate the angle α between its normal vector and the line connecting the center of mass i (i.e. the center point angle). Then determine α i (i.e., the center angle) falls into which of the m divided intervals (i.e., the center angle sub-intervals), let H α (j) represents the number (frequency) of feature points whose angle α falls into the jth interval (j=1, 2, ..., m) (i.e., the frequency of the center point angle). The statistical formula can be specifically expressed as:

[0162]

[0163] Specifically, the angle β between adjacent normal vectors (i.e., the angle frequency of normal vectors) is calculated as follows: ij (represents the normal vector angle between the i-th and j-th adjacent feature points) (i.e., the normal vector angle), judge β ij(i.e. normal vector angle) falls into which of the n β intervals (i.e. normal vector angle subintervals). Let H β (k) represents the frequency of the β angle falling into the kth interval (k = 1, 2, ..., n) (i.e., the frequency of the normal vector angle). The statistical formula can be specifically expressed as:

[0164]

[0165] Wherein, adj(i) represents the set of feature points adjacent to the i-th feature point (ie, the target feature point set).

[0166] It should be understood that the m interval frequencies H of the angle α obtained by statistics are α (j) (i.e., the frequency of the center angle) and the frequency of n intervals about the angle β H β (k) (i.e., the frequency of the normal vector angle) is combined to form the final VFH feature descriptor histogram H (i.e., the viewpoint feature histogram), which can be expressed as an M-dimensional vector (M = m + n). The specific formula can be expressed as:

[0167] H=[H α (1), H α (2), ..., H α (m), H β (1), H β (2), ..., H β (n)].

[0168] In the above embodiment, a viewpoint feature histogram is obtained by analyzing the viewpoint feature histogram of all center point angles and all normal vector angles, which enables operators to perform dimensional measurements outside the danger zone. While ensuring the measurement accuracy of the dimensions and the calculation of the deviation of the main arch rib steel pipe components, the safety of the operators is guaranteed.

[0169] Optionally, as an embodiment of the present invention, the process of performing registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data includes:

[0170] Performing coarse registration processing on all the target feature points according to the NDT registration algorithm and the viewpoint feature histogram feature descriptor to obtain a plurality of coarsely registered point cloud data;

[0171] The ICP fine registration algorithm is used to perform fine registration processing on all the coarsely registered point cloud data to obtain multiple target point cloud data.

[0172] It should be understood that the point cloud data of the complete main arch rib steel pipe component (ie, target point cloud data) is obtained by using the least squares optimized point-to-surface ICP precise registration.

[0173] Specifically, the NDT registration algorithm primarily represents the distribution of a point cloud dataset using a probability density function. This function is optimized using the Newton iterative optimization algorithm to minimize its probability distribution. This algorithm then obtains the transformation matrix between the two point clouds. The target point cloud is then divided into a grid with at least five points per grid, achieving point cloud meshing. This method calculates the transformation parameter T, and when the maximum number of iterations is reached, the transformation parameter T corresponding to the minimum error is selected.

[0174] It should be understood that the precise registration algorithm of the present invention mainly adopts the ICP precise registration algorithm of point to surface optimized by least squares. The specific process is as follows:

[0175] From the target point cloud q i Search source point cloud p i , forming corresponding point pairs;

[0176] The error function value is as follows:

[0177] f(R, T) = ∑ i ((Rp i +Tq i )·n i ) 2 ,

[0178] Where R is the rotation matrix, T is the translation parameter, and p i is the source point cloud, q i is the target point cloud, n i It is the projection distance from the source point to the normal vector of the target point after transformation. The rotation and displacement are updated using the least squares method. The iteration stops when the number of iterations reaches the set maximum number of iterations or the registration error is less than the error function value.

[0179] It should be understood that the present invention uses the least squares optimized point-to-surface ICP fine registration on the basis of coarse registration to obtain the optimal rigid transformation matrix, and uses the matrix singular value decomposition method to calculate the transformation matrix. Through multiple iterations, the transformation matrix is ​​calculated so that f(R, T) = ∑((Rp i +Tq i )·n i ) 2 The minimum transformation parameter T is used to complete the point cloud transformation and achieve precise registration.

[0180] In the above embodiment, the multiple target point cloud data is obtained by registering and analyzing all target feature points according to the viewpoint feature histogram feature descriptor, which ensures the normal progress of the construction production activities, avoids any interference, guarantees the smooth progress of the construction progress, improves the data acquisition efficiency, data accuracy and the safety of the measurement operation, and at the same time, realizes the size measurement of the operating personnel outside the dangerous area, guarantees the safety of the operating personnel while ensuring the measurement size accuracy and calculating the deviation of the main arch rib steel pipe component.

[0181] Optionally, as another embodiment of the present application, the method of the present application comprises: obtaining the main arch rib steel pipe component point cloud data by using a three-dimensional laser scanner; preprocessing the point cloud component data; using a related registration algorithm to splice the point cloud scanning model data, and converting and outputting the point cloud in pcd format, and importing the CAD model file in STL format into Geomagic software; comparing the design model and the scanning model to generate an analysis comparison report. This method can realize the size measurement of the operating personnel outside the dangerous area, guarantee the safety of the operating personnel while ensuring the measurement size accuracy and calculating the deviation of the main arch rib steel pipe component.

[0182] Optionally, as another embodiment of the present application, the present application directly obtains point cloud data by using a three-dimensional laser scanner, and performs 3D comparison and analysis with a CAD model, which can realize the size measurement of the operating personnel outside the dangerous area, guarantee the safety of the operating personnel while ensuring the measurement size accuracy and calculating the deviation of the main arch rib steel pipe component.

[0183] Optionally, as another embodiment of the present application, as shown in Figure 2 The specific process of analyzing and comparing the main arch rib steel pipe component point cloud scanning data and the CAD design reference model to obtain the appearance size deviation analysis report is as follows:

[0184] In the Geomagic Control software, the main arch rib steel pipe component point cloud data is set as a control model; in the Geomagic Control software, the SolidWorks model of the main arch rib steel pipe component is set as a design model; the above design model and the scanning model are adjusted to the same position and size, and are automatically position-fitted according to the prominent features of the main arch rib steel pipe component; the scanning model and the SolidWorks design model of the main arch rib steel pipe component are compared after fitting, and the appearance size, deviation heat map and comparison deviation analysis result of the component are obtained.

[0185] The above comparison and analysis comprises the following steps:

[0186] Deviation parameters were set in the 3D analysis function of Geomagic Control software, and critical and nominal values ​​were set in the color spectrum column. The comparative analysis function of Geomagic Control software was used to calculate the deviation of the main arch rib steel pipe components, generating a deviation heat map. Geomagic Control software also automatically generated a model comparative analysis report, which included, but was not limited to, the overall dimensions and installation accuracy of the main arch rib steel pipe components. This comparative analysis report identified construction and design issues, and locations that did not meet construction standards were promptly marked. The deviations between the point cloud data of the main arch rib steel pipe components and the SolidWorks model were annotated to obtain the final analysis results.

[0187] Specifically, if Figure 2 As shown in the figure, the registered point cloud data of the main arch rib steel pipe component and the CAD design reference model are loaded into Geomagic software together, and the point cloud scanning data of the main arch rib steel pipe component and the CAD design reference model are analyzed and compared to obtain the external dimension analysis report. Specifically:

[0188] Load the aligned point cloud data of the main arch rib steel tube component and the CAD design reference model into the Geomagic software. Save the processed data of the spliced ​​and aligned point cloud data of the main arch rib steel tube component as a pcd format file. Export the SolidWorks model of the main arch rib steel tube component as an STL format file. Import the pcd format file and the STL format file into the Geomagic software to ensure that the data transmission and processing formats match. The pcd format file of the spliced ​​and aligned point cloud data of the main arch rib steel tube component is processed by Geomagic software such as faceting, and the point cloud scanning data of the main arch rib steel tube component and the CAD design reference model are analyzed and compared to obtain an external dimension analysis report. Set the point cloud data of the main arch rib steel tube component as the control model (test data) in the Geomagic Control software. Set the SolidWorks model of the main arch rib steel tube component as the design model (reference data) in the Geomagic Control software. The design model and the scanned model were adjusted to the same position and scale, and automatic position fitting was performed based on the prominent features of the main arch rib steel pipe component. This was achieved using the Geomagic Control software function. The scanned model of the main arch rib steel pipe component and the SolidWorks design model were fitted and compared to obtain the component's external dimensions, deviation heat map, and comparative analysis results. Deviation parameters were set in the 3D analysis function of the Geomagic Control software. Deviation parameters included standard deviation, average deviation, upper deviation value, and lower deviation value. Critical values ​​and nominal values ​​were set in the color spectrum column, including minimum nominal value and maximum nominal value, with the tolerance threshold set to 0.006m. The comparative analysis function of the Geomagic Control software was used to calculate the deviation of the main arch rib steel pipe component and obtain a deviation heat map. The Geomagic Control software was used to automatically generate a model comparative analysis report. The comparative analysis report included, but was not limited to, the overall dimensions of the main arch rib steel pipe component and installation accuracy. The generated comparative analysis report was used to identify construction and design issues, and the deviations between the point cloud data of the main arch rib steel pipe component and the CAD model were annotated to obtain the final analysis results.

[0189] Optionally, as another embodiment of the present invention, compared with the prior art, the advantages of the present invention are:

[0190] 1. The present invention utilizes the characteristic of the main arch rib steel tube component with a smooth surface change and proposes a feature point extraction method and a registration method obtained by calculating the sine value of the weighted vector at different scales. This method can improve the registration stability and accuracy for the main arch rib steel tube component with a large surface that changes smoothly.

[0191] 2. The present invention mainly uses a three-dimensional laser scanner to perform a comprehensive scan on the surface of the main arch rib steel pipe component, achieving complete coverage of the component surface and meeting the equipment characteristics of high precision and easy operation. Operators can complete accurate measurement of the target surface by scanning points outside the dangerous area, effectively avoiding the risks of operating in potentially dangerous construction areas, ensuring the normal progress of construction and production activities, avoiding any interference, and ensuring the smooth progress of construction progress, thereby improving data acquisition efficiency, data accuracy, and the safety of measurement operations.

[0192] 3. Using existing technology and algorithms, the accuracy can only reach 2-3cm, while the present invention can improve the accuracy to 2mm.

[0193] Optionally, as another embodiment of the present invention, the present invention mainly includes four parts: point cloud acquisition, point cloud preprocessing, point cloud registration, and comparative analysis between point cloud and design model.

[0194] 1. Point cloud acquisition stage:

[0195] Clean the surface of the main arch rib steel pipe component to avoid any obstruction between the main arch rib steel pipe component and the 3D laser scanner; because the main arch rib steel pipe component material is affected by the temperature environment deformation, the 3D laser scanner is selected to operate at night and early morning to ensure the environment is 20℃ (the allowable temperature difference is within ±2°).

[0196] Preferably, the 3D laser scanner uses the high-precision portable Austrian RIEGLVZ-1000, which can meet the needs of large-scene scanning tasks and provide complete on-site data without repeated scanning.

[0197] Determine the positional relationship between several scanning stations and the main arch rib steel pipe components based on their shape and height, ensuring that the scanning range of the 3D laser scanner at the several scanning stations can cover the main arch rib steel pipe components;

[0198] The 3D laser scanner's parameters were adjusted to high-frequency mode to reduce the impact of environmental factors on the scanning results. To prevent on-site construction personnel from passing through the scanning area and affecting the scanning results, the construction site was temporarily blocked off. During the scanning process, the 3D laser scanner's status was carefully monitored to prevent vibration. Based on the parameter adjustments, the 3D laser scanner scanned the main arch rib steel pipe components at each scanning station, obtaining point cloud scan data P and Q of the main arch rib steel pipe components.

[0199] 2. Point cloud preprocessing stage:

[0200] The point cloud of the main arch rib steel tube component from two different perspectives is downsampled and preprocessed using the nearest point voxel filtering algorithm. First, a three-dimensional voxel grid is created for the input point cloud data, and the centroid of all points in each voxel is calculated. Then, the KD-Tree algorithm is used to traverse the original point cloud to find the point closest to the centroid in the point cloud. This point is used to replace all points in the voxel to complete the downsampling preprocessing operation.

[0201] 3. Point cloud registration stage:

[0202] The present invention improves the feature point extraction in the registration stage by adding the sine value of weighted vectors at different scales to extract feature points. That is, after the point cloud data of the main arch rib steel pipe component is down-sampled by the nearest point voxel filtering algorithm, the ISS feature point extraction algorithm is used and a method based on the sine value of weighted vectors at different scales is proposed to extract the top-ranked feature points and output the final feature points. Then, according to the calculated VFH feature descriptor, a rough registration is performed to make the two point clouds obtain a suitable initial pose. Finally, the least squares optimized point-to-surface ICP fine registration is used to obtain the optimal rigid transformation matrix.

[0203] 4. Comparative analysis of point cloud and design model:

[0204] The registered point cloud data of the main arch rib steel tube components and the CAD design reference model were loaded into Geomagic software. The point cloud scanning data of the main arch rib steel tube components and the CAD design reference model were analyzed and compared to obtain a dimensional analysis report.

[0205] Figure 3 This is a module block diagram of a component deviation measurement device provided by an embodiment of the present invention.

[0206] Alternatively, as another embodiment of the present invention, Figure 3 As shown, a component deviation measuring device includes:

[0207] A scanning module is used to scan the main arch rib steel pipe component to be measured using a 3D laser scanner to obtain multiple original point cloud data;

[0208] A filtering module, configured to filter all the original point cloud data to obtain a plurality of filtered point cloud data;

[0209] A feature analysis module is used to perform feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points;

[0210] A registration analysis module, configured to perform registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data;

[0211] The measurement result acquisition module is used to import a component reference model, perform deviation measurement on all the target point cloud data through the component reference model, and obtain component deviation measurement results.

[0212] Alternatively, another embodiment of the present invention provides a component deviation measurement system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the component deviation measurement method described above is implemented. The system may be a computer or other system.

[0213] Optionally, another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the component deviation measurement method as described above is implemented.

[0214] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0215] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0216] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical functional division. In actual implementation, other division methods may be used, such as combining or integrating multiple units or components into another system, or ignoring or not implementing certain features.

[0217] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the objectives of the embodiments of the present invention.

[0218] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0219] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0220] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A component deviation measurement method, characterized in that: The steps include: The main arch rib steel pipe component to be measured is scanned by a 3D laser scanner to obtain multiple original point cloud data; Performing filtering on all the original point cloud data to obtain a plurality of filtered point cloud data; Performing feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points; Performing registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data; Importing a component reference model, performing deviation measurement on all target point cloud data using the component reference model, and obtaining component deviation measurement results; The process of performing feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and a plurality of target feature points includes: Performing feature point analysis on all the filtered point cloud data to obtain multiple target feature points; Performing feature descriptor analysis on all target feature points to obtain viewpoint feature histogram feature descriptors; The process of analyzing the feature points of all the filtered point cloud data to obtain multiple target feature points includes: Performing local feature analysis on all the filtered point cloud data to obtain a plurality of first initial feature points; A weighted vector sine value analysis is performed on all the filtered point cloud data, and the analysis results are used as the second initial feature point set, and the feature points in the first initial feature point set and the second initial feature point set are all used as target feature points, thereby obtaining multiple target feature points.

2. The component deviation measurement method according to claim 1, characterized in that: The process of performing local feature analysis on all the filtered point cloud data to obtain a plurality of first initial feature points includes: Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset first radius to obtain a plurality of first neighborhood points corresponding to each of the filtered point cloud data; The first point cloud weights are calculated for each of the filtered point cloud data and the multiple first neighborhood points corresponding to each of the filtered point cloud data using the first formula to obtain multiple first point cloud weights corresponding to each of the filtered point cloud data. The first formula is: Among them, w1 ij is the first point cloud weight corresponding to the jth first neighborhood point of the i-th filtered point cloud data, p i is the i-th filtered point cloud data, is the jth first neighborhood point corresponding to the i-th filtered point cloud data; The covariance matrix corresponding to each of the filtered point cloud data is calculated by the second formula for each of the preset first radius, each of the filtered point cloud data, a plurality of first neighborhood points corresponding to each of the filtered point cloud data, and a plurality of first point cloud weights corresponding to each of the filtered point cloud data, to obtain a covariance matrix corresponding to each of the filtered point cloud data. The second formula is: Among them, C i is the covariance matrix corresponding to the i-th filtered point cloud data, r iss The preset first radius, w1 ij is the first point cloud weight corresponding to the jth first neighborhood point of the i-th filtered point cloud data, p i is the i-th filtered point cloud data, is the jth first neighborhood point corresponding to the i-th filtered point cloud data; Performing eigenvalue decomposition on each of the covariance matrices to obtain a first initial eigenvalue corresponding to each filtered point cloud data, a second initial eigenvalue corresponding to each filtered point cloud data, and a third initial eigenvalue corresponding to each filtered point cloud data; sorting the first initial eigenvalues, the second initial eigenvalues ​​corresponding to the filtered point cloud data, and the third initial eigenvalues ​​corresponding to the filtered point cloud data in ascending order to obtain first sorted eigenvalues ​​corresponding to the filtered point cloud data, second sorted eigenvalues ​​corresponding to the filtered point cloud data, and third sorted eigenvalues ​​corresponding to the filtered point cloud data; When the first sorted eigenvalue, the second sorted eigenvalue, and the third sorted eigenvalue satisfy the third formula and the fourth formula, the filtered point cloud data corresponding to the first sorted eigenvalue is used as the first initial feature point, thereby obtaining multiple first initial feature points. The third formula is: Among them, λ i2 is the second sorted eigenvalue corresponding to the i-th filtered point cloud data, λ i3 is the third sorted eigenvalue corresponding to the i-th filtered point cloud data, and ε1 is the preset first feature threshold; The fourth formula is: Among them, λ i1 is the first sorted eigenvalue corresponding to the i-th filtered point cloud data, λ i2 is the second sorted eigenvalue corresponding to the i-th filtered point cloud data, and ε2 is the preset second feature threshold.

3. The component deviation measurement method according to claim 2, characterized in that: The process of analyzing the weighted vector sine value of all the filtered point cloud data and using the analysis result as the second initial feature point set includes: Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset second radius to obtain a plurality of second neighborhood points corresponding to each of the filtered point cloud data; Performing neighborhood searches on each of the filtered point cloud data according to a K-dimensional tree algorithm and a preset third radius to obtain a plurality of third neighborhood points corresponding to each of the filtered point cloud data; Vector calculations are performed on each of the filtered point cloud data, the multiple second neighborhood points corresponding to each of the filtered point cloud data, and the multiple third neighborhood points corresponding to each of the filtered point cloud data using a first set of equations to obtain multiple first neighborhood point vectors corresponding to each of the filtered point cloud data and multiple second neighborhood point vectors corresponding to each of the filtered point cloud data. The first set of equations is: in, is the first neighborhood point vector corresponding to the ath second neighborhood point of the i-th filtered point cloud data, p i is the i-th filtered point cloud data, is the ath second neighborhood point corresponding to the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data, is the bth third neighborhood point corresponding to the i-th filtered point cloud data; The second point cloud weights are calculated for each of the plurality of first neighborhood point vectors corresponding to each of the filtered point cloud data by using the fifth formula to obtain a plurality of second point cloud weights corresponding to each of the filtered point cloud data. The fifth formula is: Among them, w2 ia is the second point cloud weight corresponding to the a-th second neighborhood point of the i-th filtered point cloud data, is the first neighborhood point vector corresponding to the a-th second neighborhood point of the i-th filtered point cloud data; The third point cloud weights are calculated for each of the plurality of second neighborhood point vectors corresponding to each of the filtered point cloud data using the sixth formula to obtain a plurality of third point cloud weights corresponding to each of the filtered point cloud data. The sixth formula is: Among them, w3 ib is the third point cloud weight corresponding to the bth third neighborhood point of the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data; The neighborhood angles are calculated for each of the first neighborhood point vectors corresponding to each of the filtered point cloud data, the second neighborhood point vectors corresponding to each of the filtered point cloud data, the second point cloud weights corresponding to each of the filtered point cloud data, and the third point cloud weights corresponding to each of the filtered point cloud data using the seventh formula to obtain an initial neighborhood angle corresponding to each of the filtered point cloud data. The seventh formula is: Among them, θ i is the initial neighborhood angle corresponding to the i-th filtered point cloud data, is the first neighborhood point vector corresponding to the ath second neighborhood point of the i-th filtered point cloud data, is the second neighborhood point vector corresponding to the bth third neighborhood point of the i-th filtered point cloud data, w2 ia is the second point cloud weight corresponding to the ath second neighborhood point of the i-th filtered point cloud data, w3 ib is the third point cloud weight corresponding to the bth third neighborhood point of the i-th filtered point cloud data; Sorting all the initial neighborhood angles in descending order to obtain sorted neighborhood angles corresponding to the filtered point cloud data; Filtering out the sorted neighborhood angles that meet a screening condition from all the sorted neighborhood angles, and obtaining a plurality of filtered neighborhood angles after screening, wherein the screening condition is that the sorted neighborhood angle is less than a preset angle threshold; The filtered point cloud data corresponding to the first A filtered neighborhood angles are used as second initial feature points, thereby obtaining multiple second initial feature points, and all of the second initial feature points are collected to obtain a second initial feature point set.

4. The component deviation measurement method according to claim 1, characterized in that: The process of analyzing the feature descriptors of all the target feature points to obtain the viewpoint feature histogram feature descriptors includes: Using a k-nearest neighbor algorithm to perform neighborhood searches on each of the target feature points, to obtain a plurality of fourth neighborhood points corresponding to each of the target feature points; Calculating normal vectors of a plurality of fourth neighborhood points corresponding to each of the target feature points using a principal component analysis algorithm to obtain normal vectors corresponding to each of the target feature points; The center point of all the target feature points is calculated by the seventh formula to obtain the center point The seventh formula is: in, is the X-axis coordinate of the center point, is the Y-axis coordinate of the center point, is the Z-axis coordinate of the center point, N is the total number of target feature points, x k is the X-axis coordinate of the k-th target feature point, y k is the Y-axis coordinate of the k-th target feature point, z k is the Z-axis coordinate of the k-th target feature point; The center point angle corresponding to each target feature point is calculated by the eighth formula respectively for the center point, each target feature point, and the normal vector corresponding to each target feature point to obtain the center point angle corresponding to each target feature point. The eighth formula is: Among them, α k is the center point angle corresponding to the kth target feature point, o is the center point, p4 k is the kth target feature point, n k is the normal vector corresponding to the kth target feature point; The normal vector angles corresponding to the normal vectors of each target feature point and the normal vectors corresponding to any adjacent target feature point are calculated by the ninth formula to obtain multiple normal vector angles corresponding to each target feature point. The ninth formula is: Among them, β ku is the angle between the normal vectors of the kth target feature point and the uth target feature point, n k is the normal vector corresponding to the kth target feature point, n u is the normal vector corresponding to the u-th target feature point, arccos() is the inverse cosine function; Performing a viewpoint feature histogram analysis on all the center point angles and all the normal vector angles to obtain a viewpoint feature histogram; Vectorization is performed on the viewpoint feature histogram to obtain a viewpoint feature histogram feature descriptor.

5. The component deviation measurement method according to claim 4, characterized in that: The process of analyzing the viewpoint feature histogram of all the center point angles and all the normal vector angles to obtain the viewpoint feature histogram includes: Screening out the minimum value from all the center point angles, and obtaining the minimum center point angle after screening; Screening out the maximum value from all the center point angles, and obtaining the maximum center point angle after screening; Filtering out the minimum value from all the normal vector angles, and obtaining the minimum normal vector angle after filtering; Filter out the maximum value from all the normal vector angles, and obtain the maximum normal vector angle after filtering; The center point angle interval is obtained by taking the minimum center point angle as the minimum value and the maximum center point angle as the maximum value; The normal vector angle interval is obtained by taking the minimum normal vector angle as the minimum value and the maximum normal vector angle as the maximum value; Dividing the center point angle interval into a plurality of center point angle sub-intervals with equal distances; Counting the total number of all the center point angle subintervals to obtain the total number of center point angle subintervals; Dividing the normal vector angle interval into a plurality of normal vector angle subintervals at equal distances; Counting the total number of all normal vector angle subintervals to obtain the total number of normal vector angle subintervals; The frequency of the center point angle corresponding to each center point angle subinterval is calculated by performing a tenth formula on the minimum center point angle, the maximum center point angle, the total number of center point angle subintervals, and the multiple center point angles in each center point angle subinterval to obtain the center point angle frequency corresponding to each center point angle subinterval. The tenth formula is: in, Among them, H α (j′) is the frequency of the center point angle corresponding to the j′th center point angle subinterval, N is the total number of target feature points, is the indicator function, α min is the minimum center point angle, α max is the maximum center point angle, Δα is the center point angle interval width, α k is the center point angle corresponding to the k-th target feature point, m′ is the total number of center point angle subintervals, and j′ is the j′-th center point angle subinterval; The frequency of the normal vector angle corresponding to each normal vector angle subinterval is obtained by calculating the minimum normal vector angle, the maximum normal vector angle, the total number of normal vector angle subintervals, and the number of normal vector angles in each normal vector angle subinterval using the eleventh formula. The eleventh formula is: in, Among them, H β (k′) is the frequency of the normal vector angle corresponding to the k′th normal vector angle subinterval, N is the total number of target feature points, is the indicator function, adj(k) is the set of target feature points adjacent to the kth target feature point, β min is the minimum normal vector angle, Δβ is the normal vector angle interval width, β ku is the angle between the normal vectors of the kth target feature point and the uth target feature point, k′ is the k′th normal vector angle subinterval, β max is the maximum normal vector angle, n′ is the total number of normal vector angle subintervals; The viewpoint feature histogram is calculated for all the center point angle frequencies and all the normal vector angle frequencies using the twelfth formula to obtain the viewpoint feature histogram. The twelfth formula is: H=[H α (1),H α (2),…,H α (j′),H β (1),H β (2),…,H β (k′)] T , Among them, H is the viewpoint feature histogram, H α (j′) is the frequency of the central point angle corresponding to the j′th central point angle subinterval, H β (k′) is the normal vector angle frequency corresponding to the k′th normal vector angle subinterval.

6. The component deviation measurement method according to claim 1, characterized in that: The process of performing registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data includes: Performing coarse registration processing on all the target feature points according to the NDT registration algorithm and the viewpoint feature histogram feature descriptor to obtain a plurality of coarsely registered point cloud data; The ICP fine registration algorithm is used to perform fine registration processing on all the coarsely registered point cloud data to obtain multiple target point cloud data.

7. A component deviation measuring device, characterized in that: include: A scanning module is used to scan the main arch rib steel pipe component to be measured using a 3D laser scanner to obtain multiple original point cloud data; A filtering module, configured to filter all the original point cloud data to obtain a plurality of filtered point cloud data; A feature analysis module is used to perform feature analysis on all the filtered point cloud data to obtain a viewpoint feature histogram feature descriptor and multiple target feature points; A registration analysis module, configured to perform registration analysis on all the target feature points according to the viewpoint feature histogram feature descriptor to obtain a plurality of target point cloud data; A measurement result acquisition module is used to import a component reference model, perform deviation measurement on all target point cloud data through the component reference model, and obtain component deviation measurement results; The feature analysis module is specifically used for: Performing feature point analysis on all the filtered point cloud data to obtain multiple target feature points; Performing feature descriptor analysis on all target feature points to obtain viewpoint feature histogram feature descriptors; In the feature analysis module, the process of analyzing the feature points of all the filtered point cloud data to obtain multiple target feature points includes: Performing local feature analysis on all the filtered point cloud data to obtain a plurality of first initial feature points; A weighted vector sine value analysis is performed on all the filtered point cloud data, and the analysis results are used as the second initial feature point set, and the feature points in the first initial feature point set and the second initial feature point set are all used as target feature points, thereby obtaining multiple target feature points.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the component deviation measurement method according to any one of claims 1 to 6 is implemented.

Images