Deflection non-contact detection method applied to grid structure circular steel tube rod piece

Abstract

The invention discloses a deflection non-contact detection method applied to a grid structure circular steel pipe rod piece, and relates to the technical field of steel pipe rod frames. The method comprises the steps that grid structure point cloud data are acquired through a laser scanner, and high-density voxels are obtained through voxel division and density analysis screening; performing spherical fitting on the high-density voxels, and extracting spatial spherical anchor points according to a spherical fitting difference value; constructing a grid space candidate connection matrix based on the spatial spherical anchor points, screening to obtain a linear point cloud paragraph, and fitting the axis of the rod piece through a least square method based on an RANSAC (Random Sample Consensus) algorithm; a connecting line of the spatial spherical anchor points serves as a theoretical axis of the rod piece, and a deflection degree value and a deflection offset vector are calculated through offset analysis with the fitting axis; when the deflection degree value exceeds a preset threshold value, marking an abnormal rod piece, and constructing a net rack local sub-graph by taking the abnormal rod piece as a center; and calculating a deflection direction consistency index and a deflection degree consistency index in the subgraph, calculating a local risk comprehensive index by combining the two indexes, and dividing risk grades.

Description

Technical Field

[0001] The invention relates to the technical field of steel pipe racks, and in particular to a non-contact detection method for the deflection of round steel pipe rods in a grid structure. Background Art

[0002] Space trusses are widely used in buildings, bridges, industrial plants, and other fields due to their advantages such as light weight, large spans, and reasonable load-bearing properties. Circular steel tube members, as the core load-bearing components of space trusses, have a deflection that is directly related to the overall safety and stability of the structure. Over long-term use, due to factors such as load fluctuations, material aging, and environmental corrosion, circular steel tube members are prone to abnormal deflection. If not promptly inspected and evaluated, these members may cause local structural failure or even total collapse. Therefore, accurate and efficient deflection testing is critical to ensuring the safe operation of space trusses.

[0003] In the existing technology, non-contact detection of the deflection of round steel tube rods generally adopts single-point deformation measurement based on single-station laser scanning. By deploying a laser scanner at a fixed position, several key points are selected for a single round steel tube rod. The three-dimensional coordinates of these key points are obtained by scanning, and the deflection value is calculated by comparing the initial design coordinates or historical scanning data.

[0004] However, single-point deformation measurement based on single-station laser scanning avoids direct contact interference with round steel tube members, but only focuses on calculating the deflection value of a single member. It lacks a coordinated analysis of the deformation direction and degree between abnormal members and surrounding members, making it difficult to accurately assess risks from the perspective of the overall local structure. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a non-contact detection method for the deflection of round steel tube members of a grid structure to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A non-contact deflection detection method for round steel tube members of a grid structure, comprising the following steps: Step S1: Scanning the grid structure of the round steel tube rods with a laser scanner to obtain grid structure point cloud data; performing voxel division on the grid structure point cloud data to obtain grid structure voxels; performing density analysis on the grid structure voxels to obtain high-density voxels; performing spherical fitting on the high-density voxels and calculating spherical fitting difference values; screening the high-density voxels according to the spherical fitting difference values ​​to obtain spatial spherical anchor points; Step S2: Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed; by performing rod screening on the grid space candidate connection matrix, a straight point cloud segment is obtained; and by performing axis fitting on the straight point cloud segment using the least squares method based on the RANSAC algorithm, a rod fitting axis is obtained; Step S3: taking the line connecting the geometric centers of the spatial spherical anchor points as the theoretical axis of the rod, and performing an offset analysis on the theoretical axis of the rod and the fitting axis of the rod to calculate the deflection degree value and the deflection offset vector; Step S4: By comparing the deflection value of the round steel tube member with a preset threshold, if the deflection value is greater than the preset threshold, it is marked as an abnormal member; based on the abnormal member, a local subgraph of the grid is constructed; by calculating the cosine similarity of the deflection offset vector in the local subgraph of the grid, a deflection direction consistency index is obtained; and by performing a synergy analysis on the deflection value in the local subgraph of the grid, a deflection consistency index is calculated; Step S5: By combining the deflection direction consistency index and the deflection degree consistency index, a local risk comprehensive index is calculated, and risk levels are divided according to the local risk comprehensive index to achieve non-contact detection of round steel tube rods.

[0007] Preferably, scanning the grid structure of the round steel tube rods by a laser scanner to obtain grid structure point cloud data comprises the following specific steps: The entire grid structure is divided into several layers along the vertical direction. Each layer serves as an independent scanning unit and is divided into several sectors according to the distribution direction of the round steel tube members in each layer. A laser scanner is installed in the accessible area of ​​each sector in each vertical layer. In each sector, the laser scanner performs multi-angle scanning to finally obtain the grid structure point cloud data.

[0008] Preferably, the step of performing voxel division on the grid structure point cloud data to obtain grid structure voxels; and performing density analysis on the grid structure voxels to obtain high-density voxels comprises the following steps: Based on the grid structure point cloud data, the entire space is evenly gridded according to the preset voxel size to form a three-dimensional grid structure voxel. Each grid structure voxel is used as a basic spatial unit, and the number of point clouds contained in the grid structure voxel is recorded; Calculate the point cloud density in each grid structure voxel:

[0009] Among them, MD is the point cloud density in the grid structure voxel, is the number of point clouds within the voxel, is the volume of the voxel; The grid structure voxels whose point cloud density is greater than the preset density threshold are marked as high-density voxels.

[0010] Preferably, performing spherical fitting on the high-density voxels and calculating the spherical fitting difference value comprises the following steps: The high-density voxels are subjected to the least squares method. Perform spherical fitting to obtain the fitting sphere center of the mth high-density voxel and the radius of the fitting sphere ; Calculate the root mean square error of the fitted sphere for the mth high-density voxel:

[0011] in, is the root mean square error of the mth high-density voxel, is the number of point cloud points in the mth high-density voxel, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the fitting radius of the mth high-density voxel; Calculate the normal vector consistency of the fitted sphere for the mth high-density voxel:

[0012] in, is the normal vector consistency of the mth high-density voxel, std() is the standard deviation function, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the set of point cloud points in the mth high-density voxel; By combining the root mean square error and normal vector consistency, the spherical fitting difference value is calculated:

[0013] in, is the spherical fitting difference value of the mth high-density voxel, is the root mean square error of the mth high-density voxel, is the normal vector consistency of the mth high-density voxel, is the root mean square error threshold, is the normal vector consistency threshold, and are the adjustment coefficients for the root mean square error and normal vector consistency, + =1.

[0014] Preferably, the screening of high-density voxels according to the spherical fitting difference value to obtain a spatial spherical anchor point comprises the following steps: The high-density voxels are screened according to the spherical fitting difference value, and the high-density voxels with the spherical fitting difference value less than 1 are used as spatial spherical anchor points.

[0015] Preferably, the step of constructing a grid space candidate connection matrix based on the spatial spherical anchor points comprises the following specific steps: The coordinates of the center of the sphere of the i-th spatial spherical anchor point are used as the coordinates of the spatial spherical anchor point =( , , ); Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed:

[0016] in, The minimum length of the round steel tube member is 0.5m by default. The maximum length of the round steel tube member is 10m by default. is the coordinate of the i-th spatial spherical anchor point, is the coordinate of the j-th spatial spherical anchor point.

[0017] Preferably, the step of obtaining a straight line point cloud segment by performing rod screening on the candidate connection matrix of the grid space comprises the following specific steps: In the grid space connection matrix, Two spatial spherical anchor points corresponding to 1 and , calculate the central axis of the two spatial spherical anchor points , the direction vector of the central axis =( - , - , - ), the length is ;by As the center, along the direction perpendicular to the vector Direction expansion radius , forming a cylindrical search area; For K point cloud points in the cylindrical search area, count the kth point cloud point Neighborhood radius The number of points in the neighborhood , then the local density of the kth point cloud point in the cylindrical search area is: = , and calculate the gradient vector of the point cloud point by the density difference fitting method in the neighborhood , point cloud point The gradient vector With the central axis Direction vector The angle range is If the point is not in the range of Otherwise, remove the point; For the retention point , calculate the distance from the reserved point to the center axis Calculate the radial distance average and radial distance standard deviation of all retained points. , then the radial distribution of the cylindrical search area is consistent; Project all reserved points perpendicular to the central axis The plane of the 2D point is obtained, and the circle equation is fitted by the least square method to obtain the radius of the 2D point. Fit the circle and calculate the retained points Fitting error to the fitted circle :

[0018] in, For reserved points The fitting error to the fitted circle, and For reserved points The corresponding two-dimensional coordinates, a is the horizontal coordinate of the center of the fitting circle, b is the vertical coordinate of the center of the fitting circle, is the radius of the fitted circle; The circularity of the fitting circle is calculated using the fitting error:

[0019] in, is the roundness of the fitted circle, For reserved points The fitting error to the fitted circle, is the radius of the fitted circle; Will and roundness Point cloud points in a cylindrical search area of ​​0.95 mark the straight point cloud segments of the circular steel tube member.

[0020] Preferably, performing axis fitting on the straight point cloud segment by the least square method based on the RANSAC algorithm to obtain the rod fitting axis includes the following specific steps: The linear point cloud segment is fitted using the least squares method based on the RANSAC algorithm. The number of RANSAC algorithm iterations is 50, and the distance threshold is: the maximum allowable distance between the point cloud points in the linear point cloud segment and the candidate line is 2 cm. The RANSCA algorithm is iterated: two point cloud points are randomly selected from the linear point cloud segment to generate a candidate line; all points in the point cloud are traversed, and the Euclidean distance of each point to the candidate line is calculated; the number of inliers whose distance is less than the distance threshold is counted; after repeating the iteration 50 times, the candidate line with the largest number of inliers is selected as the optimal initial axis; For the set of interior points corresponding to the optimal initial axis, the least squares method is used to refit the straight line to obtain the fitting axis of the member:

[0021] in, is the coordinate of the starting point of the fitting axis, t is the length parameter, is the unit direction vector of the member fitting axis.

[0022] Preferably, the calculation of the deflection degree value and the deflection offset vector by performing an offset analysis on the theoretical axis of the rod and the fitting axis of the rod comprises the following specific steps: The parametric equation of the theoretical axis of the rod is:

[0023] in, is the theoretical axis of the rod at the i-th spatial spherical anchor point and the j-th spatial spherical anchor point, is the coordinate of the i-th spatial spherical anchor point, t is the length parameter, is the direction vector of the i-th spatial spherical anchor point and the j-th spatial spherical anchor point; Take N sampling points evenly on the fitting axis of the rod, and the nth sampling point is ,calculate Distance to the theoretical axis of the member:

[0024] in, is the nth sampling point The distance to the theoretical axis of the member, is the i-th spatial spherical anchor point and the n-th sampling point The direction vector, is the direction vector of the i-th spatial spherical anchor point and the j-th spatial spherical anchor point; Among N sampling points, select the maximum distance to the theoretical axis of the rod , as the maximum deflection of the round steel tube member, and divide the maximum deflection by the theoretical length of the theoretical axis of the member , get the normalized maximum deflection as the deflection degree value = ; According to the sampling point corresponding to the maximum deflection, the deflection offset vector of the point from the theoretical axis of the member is calculated:

[0025] in, represents the deflection offset vector, represents the coordinates of the sampling point corresponding to the maximum deflection, is the maximum value of the theoretical axis distance of the sampling point member, Indicates sampling point Coordinates of the projection point on the theoretical axis of the member.

[0026] Preferably, the calculation of the local risk comprehensive index by combining the deflection direction consistency index and the deflection degree consistency index comprises the following specific steps: The deflection direction consistency index is calculated by calculating the cosine similarity of the deflection offset vectors of all the pipe members to be matched and the abnormal members in the local subgraph of the grid:

[0027] in, is the directional consistency index of the local subgraph of the grid, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection offset vector of the bth member to be matched, is the deflection offset vector of the abnormal member; By performing a synergistic analysis on the deflection values ​​in the local subgraph of the grid, the deflection consistency index is calculated:

[0028] in, is the consistency index of deflection degree, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection value of the bth member to be matched, is the average value of the deflection degree in the local subgraph of the grid, is the deflection value of the abnormal member; By combining the directional consistency index and the deflection consistency index, the local risk comprehensive index is calculated:

[0029] in, is the comprehensive index of local risk, FZ is the consistency index of the deflection direction of the local subgraph of the grid, PX is the consistency index of the deflection degree, It is a nonlinear adjustment factor, and the default value is 1.5.

[0030] The present invention provides a non-contact deflection detection method for circular steel tube members in a grid structure, involving machine learning and deep learning technologies, which has the following beneficial effects: (1) The high-density voxels are screened according to the spherical fitting difference value to obtain the spatial spherical anchor point, which is used to accurately locate the spherical node position of the round steel tube member in the grid structure. The spatial spherical anchor point serves as a key spatial reference, providing a reliable spatial coordinate basis for the subsequent construction of the member connection relationship and determination of the theoretical axis. At the same time, it effectively eliminates the interference point cloud in the non-spherical node area, ensuring the accuracy of the subsequent member identification and axis fitting.

[0031] (2) By performing offset analysis on the theoretical axis and the fitted axis of the rod, the deflection degree value and the deflection offset vector are calculated, which aims to convert the point cloud data into a quantifiable structural deformation index: the deflection degree value can intuitively reflect the bending deformation amplitude of the rod and realize the judgment of whether a single rod is abnormal; the deflection offset vector describes the direction of deformation, which provides a directional basis for analyzing the stress state and deformation trend of the rod, and is a key bridge for the transition from geometric morphology analysis to structural performance evaluation.

[0032] (3) By combining the directional deflection consistency index and the deflection degree consistency index to calculate the local risk comprehensive index, the synergy between the abnormal member and the surrounding members in deformation direction and degree is comprehensively considered, avoiding the one-sidedness of a single indicator evaluation and reflecting the overall risk status of the local structure more comprehensively. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 This is a flowchart of the steps of a non-contact detection method for the deflection of round steel tube members in a grid structure proposed by the present invention; Figure 2 A hierarchical diagram of the steps for obtaining a deflection degree value and a deflection offset vector in a non-contact deflection detection method for circular steel tube members of a grid structure proposed by the present invention; Figure 3 This is a hierarchical diagram of the steps for obtaining a comprehensive local risk index in a non-contact deflection detection method for circular steel tube members of a grid structure proposed by the present invention. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0036] See also Figure 1-Figure 3 The present invention provides a technical solution: a non-contact deflection detection method for round steel tube members of a grid structure.

[0037] Step S1: Scan the grid structure of round steel tube rods with a laser scanner to obtain grid structure point cloud data; perform voxel division on the grid structure point cloud data to obtain grid structure voxels; perform density analysis on the grid structure voxels to obtain high-density voxels; perform spherical fitting on the high-density voxels and calculate the spherical fitting difference value; screen the high-density voxels according to the spherical fitting difference value to obtain a spatial spherical anchor point.

[0038] The entire grid structure is divided into several layers in the vertical direction. Each layer serves as an independent scanning unit. The layer interval can be determined in combination with the structure height (such as 5-10 meters) or the floor distribution to ensure the relative spatial independence of the components in each layer, which is convenient for subsequent scanning planning. Then, the circular steel tube rods in each layer are further divided into multiple sectors on the plane according to their distribution direction (such as horizontal, longitudinal, and oblique). The sector angle range is set according to the density and directional complexity of the circular steel tube rods (such as 30° to 60°) to ensure that the directions of the circular steel tube rods in the same sector are similar and to reduce the mutual occlusion of circular steel tube rods in different directions. A laser scanner is installed in the accessible area of ​​each sector in each layer in the vertical direction (such as the ground, scaffolding, temporary platform, etc.). The installation point must meet the requirement that the laser beam emitted by the laser scanner can penetrate the gap between the circular steel tube rods to cover all the areas in the target sector. The laser scanning system can scan the circular steel tube rods at multiple angles and adjust the horizontal rotation angle and vertical pitch angle to enhance the integrity of the point cloud data. Finally, the appropriate scanning parameters are set according to the size, material and distance of the circular steel tube rods, including the laser scanning resolution (point spacing), range, accuracy and frequency, etc., to balance the data volume and processing efficiency. The divided scanning units (vertical layer + sector) are scanned in sequence. During the process, the scanner position is fixed and rotated at preset angle intervals (such as 5°-10°) to cover the entire sector, record the point cloud data at each angle and ensure that the overlap of adjacent areas is ≥30%, and each unit is scanned at least 3 times to obtain redundant data. Finally, the point cloud quality is checked in real time after the scanning is completed, including the point cloud integrity of key components, whether there are blind spots and the registration accuracy. If data is missing, it is scanned again in time to finally obtain the point cloud data of the grid structure.

[0039] Based on the grid structure point cloud data, the entire space is evenly gridded according to a preset voxel size (calculated as 3-5 times the average point spacing of the entire point cloud data), forming a three-dimensional grid structure voxel. Each grid structure voxel serves as a basic spatial unit, recording the number of point clouds contained within it.

[0040] Calculate the point cloud density in each grid structure voxel:

[0041] Among them, MD is the point cloud density in the grid structure voxel, is the number of point clouds within the voxel, is the volume of the voxel.

[0042] The grid structure voxels with point cloud density greater than the preset density threshold are marked as high-density voxels. Perform spherical fitting, the objective function is to minimize the center of the sphere and radius The residual sum of squares is:

[0043] in, is the fitting sphere center of the mth high-density voxel, is the fitting radius of the mth high-density voxel, is the i-th point cloud point in the m-th high-density voxel, is the set of point cloud points in the mth high-density voxel.

[0044] Finally, the fitting sphere center of the mth high-density voxel is obtained by the least squares method. and the radius of the fitting sphere .

[0045] Calculate the root mean square error of the fitted sphere for the mth high-density voxel:

[0046] in, is the root mean square error of the mth high-density voxel, is the number of point cloud points in the mth high-density voxel, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the fitting radius of the mth high-density voxel.

[0047] Calculate the normal vector consistency of the fitted sphere for the mth high-density voxel:

[0048] in, is the normal vector consistency of the mth high-density voxel, std() is the standard deviation function, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the mth high-density voxel.

[0049] By combining the root mean square error and normal vector consistency, the spherical fitting difference value is calculated:

[0050] in, is the spherical fitting difference value of the mth high-density voxel, is the root mean square error of the mth high-density voxel, is the normal vector consistency of the mth high-density voxel, is the root mean square error threshold, is the normal vector consistency threshold, and are the adjustment coefficients for the root mean square error and normal vector consistency, + =1.

[0051] It should be noted that and They are the adjustment coefficients of the root mean square error and the normal vector consistency, respectively. When geometric accuracy is prioritized, =0.7, =0.3; when shape consistency is prioritized, it is advisable =0.4, =0.6. Root mean square error threshold Take 2.0mm as the normal vector consistency threshold Take 0.15, <1 means accept the fit. For example, when shape consistency is prioritized, When it is 1.55mm, When it is 0.12, is 0.79, indicating that the fit is acceptable.

[0052] It's important to note that the spherical fit difference is calculated by combining root mean square error (RMS) and normal consistency. RMS quantifies the average distance deviation between the point cloud and the fitted sphere, reflecting the overall geometric fit. Normal consistency, on the other hand, measures the dispersion of the point cloud's local surface orientation (i.e., normal vector) around the center of the fitted sphere, revealing whether the point cloud exhibits the uniform outward radiation characteristic of a sphere in its local topology. Using RMS error alone cannot distinguish pseudo-spheres where the point cloud is densely distributed but has chaotic orientations (such as dense pits or bumps). Using normal consistency alone ignores pseudo-fits where the point cloud as a whole is too far from the sphere. By combining these weighted approaches to form a spherical fit difference, we can more comprehensively determine whether high-density voxels truly correspond to spherical nodes (anchor points) in space. This effectively eliminates interfering areas where the point cloud is densely clustered but non-spherical in shape (such as rod parts, connecting plates, or obstructions), ensuring the accuracy of spatial spherical anchor point identification in subsequent steps.

[0053] High-density voxels are screened according to the spherical fitting difference value, and high-density voxels with a spherical fitting difference value less than 1 are used as spatial spherical anchor points.

[0054] Step S2: Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed; by performing rod screening on the grid space candidate connection matrix, a straight point cloud segment is obtained; and by performing axis fitting on the straight point cloud segment using the least squares method based on the RANSAC algorithm, a rod fitting axis is obtained.

[0055] The coordinates of the center of the sphere of the i-th spatial spherical anchor point are used as the coordinates of the spatial spherical anchor point =( , , ).

[0056] Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed:

[0057] in, The minimum length of the round steel tube member is 0.5m by default. The maximum length of the round steel tube member is 10m by default. is the coordinate of the i-th spatial spherical anchor point, is the coordinate of the j-th spatial spherical anchor point.

[0058] In the grid space connection matrix, for Two spatial spherical anchor points corresponding to 1 and , calculate the central axis of the two spatial spherical anchor points , the direction vector of the central axis =( - , - , - ), the length is .by As the center, along the direction perpendicular to the vector Direction expansion radius (Take 1.5-2 times the maximum design diameter of the round steel pipe rod) to form a cylindrical search area.

[0059] For each point cloud point in the cylindrical search area , through the neighborhood radius of the point cloud point (5cm) neighborhood, count the number of points in the neighborhood , then the local density of the kth point cloud point in the cylindrical search area is: = , and establish a local linear density field model by fitting the density difference within the neighborhood (e)= + * , ( ) represents the first Local density of neighborhood points: Taking the e-th neighborhood point as the center, calculate the neighborhood radius of the e-th neighborhood point The local density within Represents the e-th neighborhood point and point cloud point in the neighborhood The distance is finally solved by minimizing the weighted residual square sum to obtain the gradient vector of the k-th point cloud point ), get the gradient vector of the point cloud point , if the gradient vector at this point Direction vector with respect to the central axis The angle is close to 90° (90° ), then the point is retained; otherwise, the point is discarded.

[0060] For the retention point , calculate the distance from this point to the center axis The radial distance of all retained points is calculated, and the radial distance mean and radial distance standard deviation are calculated. , then the radial distribution is considered to be consistent (which is consistent with the characteristic of stable radius of the steel pipe section).

[0061] Project all reserved points perpendicular to the central axis The plane (take the cross section in the middle of the central axis) is used to obtain the corresponding two-dimensional points. All the corresponding two-dimensional points are fitted with the circle equation by the least square method to obtain the radius of Fit the circle and calculate the retained points Fitting error to the fitted circle :

[0062] in, For reserved points The fitting error to the fitted circle, and For reserved points The corresponding two-dimensional coordinates, a is the horizontal coordinate of the center of the fitting circle, b is the vertical coordinate of the center of the fitting circle, is the radius of the fitted circle.

[0063] The circularity of the fitting circle is calculated using the fitting error:

[0064] in, is the roundness of the fitted circle, For reserved points The fitting error to the fitted circle, is the radius of the fitted circle.

[0065] Will and roundness Point cloud points in a cylindrical search area of ​​0.95 mark the straight point cloud segments of the circular steel tube member.

[0066] The linear point cloud segments are fitted using the least squares method based on the RANSAC algorithm. The RANSAC algorithm parameters need to be adjusted based on the steel pipe characteristics and scanning accuracy: the number of iterations is 50 to 100. Too few iterations may miss the optimal line, while too many iterations increase the computational effort. 50 iterations strike a balance between covering enough candidate lines and computational efficiency (suitable for medium-sized point clouds, approximately 1,000 to 5,000 points). Distance threshold: This sets the maximum allowable distance from a point to a line (the inlier determination criterion). The value can be "circular steel pipe rod radius error + scanning accuracy" (e.g., 2 cm). Circular steel pipe rod manufacturing has slight radius errors, usually ≤1 mm. Three-dimensional scanning errors, such as laser scanning accuracy, are approximately 1 to 2 mm. After superposition, 2 cm can cover more than 95% of inliers while excluding significantly deviated noise points.

[0067] The RANSCA algorithm is iterated: two points are randomly selected from the line point cloud segment (two points in space define a line) to generate a candidate line. All points in the point cloud are traversed, and the Euclidean distance from each point to the candidate line is calculated. The number of inliers with a distance less than a distance threshold is counted. The more inliers, the more likely the candidate line is to be the true axis. After 50 iterations, the candidate line with the largest number of inliers is selected as the optimal initial axis. The projected coordinates of all points in the inlier set along the axis are calculated, and the minimum and maximum values ​​of the projected coordinates are taken as the valid range of the fitted axis.

[0068] For the set of interior points corresponding to the optimal initial axis, the least squares method is used to refit the straight line. The least squares method further optimizes the straight line parameters (starting point, direction vector) by minimizing the sum of the squares of the distances from all interior points to the straight line, eliminating the errors caused by random sampling of the RANSAC algorithm, and obtaining a more accurate rod fitting axis:

[0069] in, is the coordinate of the starting point of the fitting axis, t is the length parameter, is the unit direction vector of the member fitting axis.

[0070] Step S3: The line connecting the geometric centers of the spatial spherical anchor points is used as the theoretical axis of the rod, and the deflection degree value and the deflection offset vector are calculated by performing an offset analysis on the theoretical axis of the rod and the fitting axis of the rod.

[0071] With the spatial anchor points of the spherical nodes at both ends of the rod as endpoints, a straight line is constructed as the theoretical axis of the rod. The parametric equation of the theoretical axis of the rod is:

[0072] in, is the theoretical axis of the rod at the i-th spatial spherical anchor point and the j-th spatial spherical anchor point, is the coordinate of the i-th spatial spherical anchor point, t is the length parameter, is the direction vector between the i-th spatial spherical anchor point and the j-th spatial spherical anchor point.

[0073] Take N sampling points evenly on the fitting axis of the rod (such as 1 point every 10 cm), and the nth sampling point is ,calculate Distance to the theoretical axis of the member:

[0074] in, is the nth sampling point The distance to the theoretical axis of the member, is the i-th spatial spherical anchor point and the n-th sampling point The direction vector, is the direction vector between the i-th spatial spherical anchor point and the j-th spatial spherical anchor point.

[0075] It should be noted that N sampling points are evenly taken on the fitting axis of the rod and the distance from each point to the theoretical axis of the rod is calculated. The purpose is to cover the entire length of the rod by multiple points, avoid the randomness of single-point measurement, and fully capture the offset of different positions of the rod, so as to accurately determine the value and location of the maximum deflection, and provide a reliable basis for the subsequent calculation of the normalized deflection value.

[0076] Among N sampling points, select the maximum distance to the theoretical axis of the rod , as the maximum deflection of the round steel tube member, and divide the maximum deflection by the theoretical length of the theoretical axis of the member , get the normalized maximum deflection as the deflection degree value = .

[0077] It should be noted that the deflection value , selecting the maximum value of the distance to the theoretical axis of the rod among N sampling points as the maximum deflection can accurately capture the most significant position and degree of the bending deformation of the rod and reflect its critical state of stress deformation; and normalizing the maximum deflection by dividing it by the theoretical length of the theoretical axis of the rod can eliminate the incomparable problem of absolute deflection values ​​caused by length differences of different rods, so that the deformation degree of round steel tube rods of different specifications and lengths has a unified measurement standard. The obtained deflection degree value can more objectively reflect the relative deformation amplitude of the rod, and provide a quantitative basis for subsequent comparison with the preset threshold to determine whether it is an abnormal rod.

[0078] According to the sampling point corresponding to the maximum deflection, the deflection offset vector of the point from the theoretical axis of the member is calculated:

[0079] in, represents the deflection offset vector, represents the coordinates of the sampling point corresponding to the maximum deflection, is the maximum value of the theoretical axis distance of the sampling point member, Indicates sampling point Coordinates of the projection point on the theoretical axis of the member.

[0080] In order to prevent error interference in a single detection, a multi-frame scanning fusion strategy is further introduced, that is, by repeatedly fitting the same rod at multiple time points or multiple angles, the average value of the deflection degree value and the average value of the deflection offset vector are calculated as the final result output.

[0081] Step S4: By comparing the deflection value of the round steel tube member with a preset threshold, if it is greater than the preset threshold, it is marked as an abnormal member; based on the abnormal member, a local subgraph of the truss is constructed; by calculating the cosine similarity of the deflection offset vector in the local subgraph of the truss, the deflection direction consistency index is obtained; by performing a synergistic analysis on the deflection value in the local subgraph of the truss, the deflection consistency index is calculated.

[0082] The spatial spherical anchor points are used as node units in the grid structure topology diagram, and the rod fitting axes are used as edge units in the grid structure topology diagram, where each edge connects two node units.

[0083] Topology verification: Verify the accuracy of the graphical model through point cloud data to ensure that the connection relationship between all nodes and rods is consistent with the actual structure (such as avoiding missed connections and wrong connections, especially checking multi-rod connections at complex nodes).

[0084] Screening of large deflection members: A deflection threshold is preset (set through engineering experience), and round steel tube members with a deflection greater than the preset threshold are marked as "abnormal members".

[0085] The scope of the local subgraph of the truss is defined as follows: the spatial spherical anchor points at both ends of the abnormal member are determined as the core nodes of the local subgraph of the truss; based on this core node, the local subgraph of the truss is expanded: all circular steel tube members directly connected to the core node, as well as the spatial spherical anchor points connected to the other end of the circular steel tube member, are included in the local subgraph of the truss; finally, after the local subgraph of the truss is constructed, the circular steel tube members in the local subgraph of the truss except the abnormal member are recorded as the members to be matched.

[0086] The directional consistency index of the deflection offset vectors of all the pipe members to be matched and the abnormal members in the local subgraph of the grid is calculated by cosine similarity:

[0087] in, is the directional consistency index of the local subgraph of the grid, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection offset vector of the bth member to be matched, is the deflection offset vector of the abnormal member.

[0088] By performing a synergistic analysis on the deflection values ​​in the local subgraph of the grid, the deflection consistency index is calculated:

[0089] in, is the consistency index of deflection degree, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection value of the bth member to be matched, is the average value of the deflection degree in the local subgraph of the grid, is the deflection value of the abnormal member.

[0090] It should be noted that the deflection consistency index is obtained by introducing the average deflection value in the local subgraph of the grid. As a benchmark, quantify the mechanical synergy strength of the abnormal rod and the surrounding rods to be matched in terms of deflection value, and calculate the deviation of the rods to be matched Deviation from abnormal member The product and sum of the two reveal the correlation between the deformation trends; the denominator is standardized to eliminate the dimension effect, so that the output value reflects the consistency of the deflection degree of local structural deformation.

[0091] Step S5: By combining the deflection direction consistency index and the deflection degree consistency index, a local risk comprehensive index is calculated, and risk levels are divided according to the local risk comprehensive index to achieve non-contact detection of round steel tube rods.

[0092] By combining the directional consistency index and the deflection consistency index, the local risk comprehensive index is calculated:

[0093] in, is the comprehensive index of local risk, FZ is the consistency index of the deflection direction of the local subgraph of the grid, PX is the consistency index of the deflection degree, It is a nonlinear adjustment factor, and the default value is 1.5.

[0094] It should be noted that the local risk comprehensive index is calculated by combining the deflection direction consistency index and the deflection degree consistency index. The deflection direction consistency index reflects the spatial correlation between the abnormal member and the surrounding members in the deformation direction, while the deflection degree consistency index quantifies their statistical correlation in deformation amplitude. CCI combines the deflection direction consistency index and the deflection degree consistency index through a nonlinear formula, in which the adjustment factor The contribution of deflection consistency has been enhanced. The local risk comprehensive index indicates that CCI outputs a high risk value only when both direction and degree show significant synergy, effectively eliminating local noise interference and reducing the misjudgment rate.

[0095] The risk levels are divided according to the local risk comprehensive index (CCI) as follows: when CCI≤0.3, it is judged to be a low risk level, indicating that the risk correlation between the deflection of the abnormal rod and the surrounding rods is weak, and the local stability of the structure is good; when 0.3<CCI≤0.7, it is judged to be a medium risk level, indicating that there is a certain synergy in the risks of the abnormal rod and the surrounding rods, and it is necessary to pay attention to the stress state of the local structure and monitor it in time; when CCI>0.7, it is judged to be a high risk level, which means that the deflection direction and degree of the abnormal rod are highly consistent with those of the surrounding rods, and the local structure may have overall stress abnormalities, and detailed inspection and reinforcement treatment must be carried out immediately.

[0096] This paper introduces a non-contact deflection detection method for circular steel tube members in grid structures. The method acquires point cloud data by vertically layering laser scanners. Spatial spherical anchor points are selected through voxel partitioning and density analysis. A connection matrix is ​​constructed based on the anchor points and the member axis is fitted. The deflection parameters are then obtained through offset analysis between the theoretical axis and the fitted axis. Finally, a comprehensive local risk index is calculated by combining the local direction and degree consistency index to achieve risk level classification and non-contact detection of the members.

[0097] High-density voxels are screened according to the spherical fitting difference values ​​to obtain spatial spherical anchor points. Their significance lies in accurately locating the spherical node positions of circular steel tube members in the truss structure. These anchor points serve as key spatial benchmarks, providing a reliable spatial coordinate basis for the subsequent construction of member connection relationships and determination of theoretical axes. At the same time, they effectively eliminate interfering point clouds in non-spherical node areas, ensuring the accuracy of subsequent member identification and axis fitting.

[0098] The deflection degree value and deflection offset vector are calculated by performing an offset analysis on the theoretical axis and the fitted axis of the rod. Its significance lies in converting point cloud data into quantifiable structural deformation indicators: the deflection degree value can intuitively reflect the bending deformation amplitude of the rod and realize the judgment of whether a single rod is abnormal; the deflection offset vector describes the direction of deformation, providing a directional basis for analyzing the stress state and deformation trend of the rod, and is a key bridge in the transition from geometric morphology analysis to structural performance evaluation.

[0099] The local risk comprehensive index is calculated by combining the deflection direction consistency index and the deflection degree consistency index. Its significance lies in comprehensively considering the synergy between the abnormal rods and the surrounding rods in the deformation direction and degree, avoiding the one-sidedness of single indicator evaluation, and more comprehensively reflecting the overall risk status of the local structure. It provides a scientific quantitative basis for accurately dividing the risk level and judging whether there is overall stress anomaly in the local structure, thereby laying the foundation for formulating targeted detection and reinforcement measures.

[0100] 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 "include," "comprise," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations. The phrase "includes an element defined by..." does not exclude the presence of other identical elements in the process, method, article, or device that includes the element.

[0101] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A non-contact deflection detection method for circular steel tube members in a grid structure, characterized by: The following steps are involved: Step S1: Scanning the grid structure of the round steel tube rods with a laser scanner to obtain grid structure point cloud data; performing voxel division on the grid structure point cloud data to obtain grid structure voxels; performing density analysis on the grid structure voxels to obtain high-density voxels; performing spherical fitting on the high-density voxels and calculating spherical fitting difference values; screening the high-density voxels according to the spherical fitting difference values ​​to obtain spatial spherical anchor points; Step S2: Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed; by performing rod screening on the grid space candidate connection matrix, a straight point cloud segment is obtained; and by performing axis fitting on the straight point cloud segment using the least squares method based on the RANSAC algorithm, a rod fitting axis is obtained; Step S3: taking the line connecting the geometric centers of the spatial spherical anchor points as the theoretical axis of the rod, and performing an offset analysis on the theoretical axis of the rod and the fitting axis of the rod to calculate the deflection degree value and the deflection offset vector; Step S4: By comparing the deflection value of the round steel tube member with a preset threshold, if the deflection value is greater than the preset threshold, it is marked as an abnormal member; based on the abnormal member, a local subgraph of the grid is constructed; by calculating the cosine similarity of the deflection offset vector in the local subgraph of the grid, a deflection direction consistency index is obtained; by performing a synergy analysis on the deflection value in the local subgraph of the grid, a deflection consistency index is calculated; Step S5: By combining the deflection direction consistency index and the deflection degree consistency index, a local risk comprehensive index is calculated, and risk levels are divided according to the local risk comprehensive index to achieve non-contact detection of round steel tube rods.

2. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 1, characterized in that: Scanning the grid structure of the round steel tube rods by a laser scanner to obtain grid structure point cloud data includes the following specific steps: The entire grid structure is divided into several layers along the vertical direction. Each layer serves as an independent scanning unit and is divided into several sectors according to the distribution direction of the round steel tube members in each layer. A laser scanner is installed in the accessible area of ​​each sector in each vertical layer. In each sector, the laser scanner performs multi-angle scanning to finally obtain the grid structure point cloud data.

3. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 2, characterized in that: The grid structure voxels are obtained by performing voxel division on the grid structure point cloud data; Performing density analysis on the grid structure voxels to obtain high-density voxels includes the following steps: Based on the grid structure point cloud data, the entire space is evenly gridded according to the preset voxel size to form a three-dimensional grid structure voxel. Each grid structure voxel is used as a basic spatial unit, and the number of point clouds contained in the grid structure voxel is recorded; Calculate the point cloud density in each grid structure voxel: ； Among them, MD is the point cloud density in the grid structure voxel, is the number of point clouds within the voxel, is the volume of the voxel; The grid structure voxels whose point cloud density is greater than the preset density threshold are marked as high-density voxels.

4. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 3, characterized in that: The method of performing spherical fitting on the high-density voxels and calculating the spherical fitting difference value comprises the following steps: The high-density voxels are subjected to the least squares method. Perform spherical fitting to obtain the fitting sphere center of the mth high-density voxel and the radius of the fitting sphere ; Calculate the root mean square error of the fitted sphere for the mth high-density voxel: ； in, is the root mean square error of the mth high-density voxel, is the number of point cloud points in the mth high-density voxel, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the fitting radius of the mth high-density voxel; Calculate the normal vector consistency of the fitted sphere for the mth high-density voxel: ； in, is the normal vector consistency of the mth high-density voxel, std() is the standard deviation function, is the i-th point cloud point in the m-th high-density voxel, is the fitting sphere center of the mth high-density voxel, , is the set of point cloud points in the mth high-density voxel; By combining the root mean square error and normal vector consistency, the spherical fitting difference value is calculated: ； in, is the spherical fitting difference value of the mth high-density voxel, is the root mean square error of the mth high-density voxel, is the normal vector consistency of the mth high-density voxel, is the root mean square error threshold, is the normal vector consistency threshold, and are the adjustment coefficients for the root mean square error and normal vector consistency, + =1.

5. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 4, characterized in that: The method of screening high-density voxels according to the spherical fitting difference value to obtain a spatial spherical anchor point includes the following steps: The high-density voxels are screened according to the spherical fitting difference value, and the high-density voxels with the spherical fitting difference value less than 1 are used as spatial spherical anchor points.

6. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 5, characterized in that: The step of constructing a grid space candidate connection matrix based on the spatial spherical anchor points comprises the following specific steps: The coordinates of the center of the sphere of the i-th spatial spherical anchor point are used as the coordinates of the spatial spherical anchor point =( , , ); Based on the spatial spherical anchor points, a grid space candidate connection matrix is ​​constructed: ； in, The minimum length of the round steel tube member is 0.5m by default. The maximum length of the round steel tube member is 10m by default. is the coordinate of the i-th spatial spherical anchor point, is the coordinate of the j-th spatial spherical anchor point.

7. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 6, characterized in that: The method of obtaining a straight line point cloud segment by screening the candidate connection matrix of the grid space includes the following specific steps: In the grid space connection matrix, Two spatial spherical anchor points corresponding to 1 and , calculate the central axis of the two spatial spherical anchor points , the direction vector of the central axis =( - , - , - ), the length is ;by As the center, along the direction perpendicular to the vector Direction expansion radius , forming a cylindrical search area; For K point cloud points in the cylindrical search area, count the kth point cloud point Neighborhood radius The number of points in the neighborhood , then the local density of the kth point cloud point in the cylindrical search area is: = , and calculate the gradient vector of the point cloud point by the density difference fitting method in the neighborhood , point cloud point The gradient vector With the central axis Direction vector The angle range is If the point is not in the range of Otherwise, remove the point; For the retention point , calculate the distance from the reserved point to the center axis Calculate the radial distance average and radial distance standard deviation of all retained points. , then the radial distribution of the cylindrical search area is consistent; Project all reserved points perpendicular to the central axis The plane of the 2D point is obtained, and the circle equation is fitted by the least square method to obtain the radius of the 2D point. The fitting circle and calculation of the retained points Fitting error to the fitted circle : ； in, For reservation point The fitting error to the fitted circle, and For reservation point The corresponding two-dimensional coordinates, a is the horizontal coordinate of the center of the fitting circle, b is the vertical coordinate of the center of the fitting circle, is the radius of the fitted circle; The circularity of the fitting circle is calculated using the fitting error: ； in, is the roundness of the fitted circle, For reservation point The fitting error to the fitted circle, is the radius of the fitted circle; Will and roundness Point cloud points in a cylindrical search area of ​​0.95 mark the straight point cloud segments of the circular steel tube member.

8. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 7, characterized in that: The axis fitting of the straight point cloud segment by the least square method based on the RANSAC algorithm to obtain the rod fitting axis includes the following specific steps: The linear point cloud segment is fitted using the least squares method based on the RANSAC algorithm. The number of RANSAC algorithm iterations is 50, and the distance threshold is: the maximum allowable distance between the point cloud points in the linear point cloud segment and the candidate line is 2 cm. The RANSCA algorithm is iterated: two point cloud points are randomly selected from the linear point cloud segment to generate a candidate line; all points in the point cloud are traversed, and the Euclidean distance of each point to the candidate line is calculated; the number of inliers whose distance is less than the distance threshold is counted; after repeating the iteration 50 times, the candidate line with the largest number of inliers is selected as the optimal initial axis; For the set of interior points corresponding to the optimal initial axis, the least squares method is used to refit the straight line to obtain the fitting axis of the member: ； in, is the coordinate of the starting point of the fitting axis, t is the length parameter, is the unit direction vector of the member fitting axis.

9. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 8, characterized in that: The method of calculating the deflection value and the deflection offset vector by performing an offset analysis on the theoretical axis of the rod and the fitting axis of the rod includes the following specific steps: The parametric equation of the theoretical axis of the rod is: ； in, is the theoretical axis of the rod at the i-th spatial spherical anchor point and the j-th spatial spherical anchor point, is the coordinate of the i-th spatial spherical anchor point, t is the length parameter, is the direction vector of the i-th spatial spherical anchor point and the j-th spatial spherical anchor point; Take N sampling points evenly on the fitting axis of the rod, and the nth sampling point is ,calculate Distance to the theoretical axis of the member: ； in, is the nth sampling point The distance to the theoretical axis of the member, is the i-th spatial spherical anchor point and the n-th sampling point The direction vector, is the direction vector of the i-th spatial spherical anchor point and the j-th spatial spherical anchor point; Among N sampling points, select the maximum distance to the theoretical axis of the rod , as the maximum deflection of the round steel tube member, and divide the maximum deflection by the theoretical length of the theoretical axis of the member , get the normalized maximum deflection as the deflection degree value = ; According to the sampling point corresponding to the maximum deflection, the deflection offset vector of the point from the theoretical axis of the member is calculated: ； in, represents the deflection offset vector, represents the coordinates of the sampling point corresponding to the maximum deflection, is the maximum value of the theoretical axis distance of the sampling point member, Indicates sampling point Coordinates of the projection point on the theoretical axis of the member.

10. The non-contact deflection detection method for round steel tube members of a grid structure according to claim 9, characterized in that: The calculation of the local risk comprehensive index by combining the deflection direction consistency index and the deflection degree consistency index includes the following specific steps: The deflection direction consistency index is calculated by calculating the cosine similarity of the deflection offset vectors of all the pipe members to be matched and the abnormal members in the local subgraph of the grid: ； in, is the directional consistency index of the local subgraph of the grid, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection offset vector of the bth member to be matched, is the deflection offset vector of the abnormal member; By performing a synergistic analysis on the deflection values ​​in the local subgraph of the grid, the deflection consistency index is calculated: ； in, is the consistency index of deflection degree, B is the number of rods to be matched in the local subgraph of the grid, b is the index of the rod to be matched, is the deflection value of the bth member to be matched, is the average value of the deflection degree in the local subgraph of the grid, is the deflection value of the abnormal member; By combining the directional consistency index and the deflection consistency index, the local risk comprehensive index is calculated: ； in, is the comprehensive index of local risk, FZ is the consistency index of the deflection direction of the local subgraph of the grid, PX is the consistency index of the deflection degree, It is a nonlinear adjustment factor, and the default value is 1.5.

Images