A method for detecting surface flatness of precast beam segments based on three-dimensional point cloud model

Through coordinate calibration and normal vector analysis of the three-dimensional point cloud model, the problems of low flatness detection efficiency and low accuracy of prefabricated beam sections are solved, and fast and accurate flatness detection is achieved, which improves the degree of automation and digital integration.

CN114037706BActive Publication Date: 2024-11-12SOUTHEAST UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202111200770.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-13
Publication Date
2024-11-12
Estimated Expiration
2041-10-13

AI Technical Summary

Technical Problem

The prior art has low detection efficiency and low accuracy in the flatness detection of prefabricated beam sections, and the results are difficult to digitally manage, so the surface flatness information cannot be extracted quickly and accurately.

Method used

Using a detection method based on the three-dimensional point cloud model, the coordinate calibration, normal vector extraction and optimal reference surface determination of the three-dimensional point cloud model is carried out to achieve rapid and accurate detection of the surface flatness of the prefabricated beam section. Specific steps include coarse calibration, precision calibration, normal vector analysis and iterative search for the optimal reference surface and calculating flatness.

Benefits of technology

Improves computing efficiency and automation, avoids errors from traditional contact measurements, saves calculation time, and improves the digital integration of detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114037706B_ABST
    Figure CN114037706B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for detecting the surface flatness of a prefabricated beam segment based on a three-dimensional point cloud model, comprising the following steps: (1) according to the specific form of the three-dimensional point cloud model of a target component in a three-dimensional rectangular coordinate system, the model is coarsely calibrated and finely calibrated in turn, a spatial rotation matrix is ​​determined, and point cloud coordinate calibration is performed; (2) according to the principle of principal component analysis and K nearest neighbor method, the normal vector at each point position of the component three-dimensional point cloud model is determined, and by limiting the direction and coordinate interval of the normal vector, segmentation and extraction of a certain surface to be measured are achieved; (3) according to the morphological relationship between the surface to be measured and the three-dimensional rectangular coordinate system, an optimal reference surface is iteratively searched and the flatness of the surface is calculated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of manufacturing quality inspection of prefabricated beam segments in bridge engineering, and in particular to a method for inspecting the surface flatness of prefabricated beam segments based on a three-dimensional point cloud model. Background Art

[0002] Flatness is an important evaluation index for the manufacturing quality inspection of prefabricated beam segments. It not only reflects the beauty of the construction effect, but also has a certain impact on the mechanical properties of the prefabricated beam segments during service. Flatness that exceeds the specification limit also represents the unreliable manufacturing quality of the prefabricated beam segments to a certain extent. At present, the detection method for the flatness of prefabricated beam segments in actual projects usually adopts contact measurement methods such as feeler gauges and profilometers. The detection efficiency is low and the accuracy is not high. The detection results are not easy to manage digitally. The three-dimensional point cloud model is a collection of massive points containing the characteristics of the target surface. The spatial form of the object surface is reflected based on the three-dimensional coordinates of each point. The surface feature information of the prefabricated component is collected by methods such as three-dimensional laser scanners. Compared with traditional detection methods, it has the advantages of non-contact, high sampling rate, and high degree of automation. However, based on the point cloud model containing massive three-dimensional coordinate information data, how to quickly extract the target surface from the overall model, how to accurately realize the flatness detection of the surface, and how to integrate it into the automated calculation steps are urgent to solve. Summary of the invention

[0003] Purpose of the invention: For prefabricated beam segments with relatively regular geometric shapes, a surface flatness detection method based on a three-dimensional point cloud model is proposed, which is easy to program and implement, and improves the calculation efficiency and degree of automation compared with traditional detection methods.

[0004] Technical solution: A method for detecting the surface flatness of a prefabricated beam segment based on a three-dimensional point cloud model, the method comprising the following steps:

[0005] (1) According to the specific form of the 3D point cloud model of the target component in the 3D rectangular coordinate system, the model is coarsely calibrated and finely calibrated in sequence, the spatial rotation matrix is ​​determined, and the point cloud coordinate calibration is performed;

[0006] (2) According to the principle of principal component analysis and K nearest neighbor method, the normal vector at each point of the component 3D point cloud model is determined, and the segmentation and extraction of a certain surface to be measured is realized by limiting the direction and coordinate range of the normal vector;

[0007] (3) According to the morphological relationship between the surface to be measured and the three-dimensional rectangular coordinate system, the optimal reference surface is iteratively found and the flatness of the surface is calculated.

[0008] Preferably, steps (1), (2) and (3) all need to satisfy the constraints: the geometric shape of the precast beam segment is regular, the placement position of the precast beam segment is horizontal, the origin of the three-dimensional coordinate system is set at the centroid of the point cloud model, and the X-axis, Y-axis and Z-axis directions of the three-dimensional coordinate system are respectively parallel to the component beam width, beam length and beam height directions.

[0009] Preferably, the step (1) specifically comprises the following steps:

[0010] 1.1 For the 3D point cloud model Pt0 of the precast beam segment, set the origin O of the original coordinate system X0Y0Z0 to the centroid position of the 3D point cloud model:

[0011]

[0012] Among them, Pt c is the 3D point cloud model after centralization; x1...x n is the coordinate value of each point X0 in Pt0, y1...y n is the Y0 coordinate value, z1...z n is the Z0 coordinate value; μ X is the average value of the X0 coordinates of each point in Pt0, μ Y is the average value of the original Y0 coordinate value, μ Z is the average value of the original Z0 coordinate value;

[0013] 1.2 For the horizontally placed precast beam segment, determine that the initial Z1 coordinate axis positive direction is the opposite direction of gravity and parallel to the beam height direction, and make a three-dimensional point cloud model Pt c Project on the X0OY0 plane and perform principal component analysis on the projection. According to the principle of principal component analysis, firstly, the projected point cloud is decentralized, the covariance matrix of the decentralized point cloud is calculated, and the covariance matrix is ​​decomposed by singular value to obtain a set of eigenvalues ​​and eigenvectors uniquely corresponding to each eigenvalue. The eigenvector corresponding to the largest eigenvalue is the first principal component, and the eigenvector corresponding to the second largest eigenvalue is the second principal component. The Y1 coordinate axis direction is defined as the first principal component direction, and the X1 coordinate axis direction is defined as the second principal component direction. The calibration of the initial coordinate system X1Y1Z1 is completed, that is, the rough calibration of the coordinates of the three-dimensional point cloud model.

[0014] 1.3 Slices are made at appropriate positions that can reflect the characteristics of beam width, beam length, and beam height of the component. The slice thickness is twice the point cloud density. The point cloud contained in the slice is projected onto the slice plane and principal component analysis is performed. The projection is fitted into a straight line. The angles between the projection line and the two coordinate axes are compared. According to the principle of principal component analysis, the direction of the coordinate axis with the smallest angle with the projection line is defined as the first principal component direction, and the direction of the other coordinate axis in the slice plane is defined as the second principal component direction. After completing the above steps in the beam width, beam length, and beam height directions, the calibration of the final coordinate system XYZ is completed, that is, the precise calibration of the coordinates of the three-dimensional point cloud model.

[0015] According to the angle between the final coordinate system XYZ and the initial coordinate system X1Y1Z1, calculate the rotation matrix R corresponding to the change of the point cloud from the initial coordinate system X1Y1Z1 to the final coordinate system XYZ x , R y , R z , if the rotation direction from the initial coordinate system X1Y1Z1 to the final coordinate system XYZ is counterclockwise, then the rotation matrix R is obtained x , R y , R z They are:

[0016]

[0017] If the rotation direction is clockwise, the rotation matrix R x , R y , R z They are:

[0018]

[0019] Among them, α is the angle between the coordinate system Y1OZ1 and the coordinate system YOZ; β is the angle between the coordinate system X1OZ1 and the coordinate system XOZ, and γ is the angle between the coordinate system X1OZ1 and the coordinate system XOZ. According to the above R x , R y , R z Get the spatial rotation matrix R = R x ·R y ·R z , and then realize the point cloud coordinate calibration:

[0020] Pt=R·Pt c

[0021] Among them, Pt c is the centered three-dimensional point cloud model, and Pt is the three-dimensional point cloud model after coordinate calibration.

[0022] Preferably, the step (2) specifically comprises the following steps:

[0023] 2.1 According to the K nearest neighbor method, for the i-th point p in the 3D point cloud model Pt i (x i ,y i , z i ), select the nearest K neighboring points:

[0024]

[0025] Among them, N is the number of points in the 3D point cloud model Pt, is the distance from point p i The set of the nearest K neighboring points, is the kth neighboring point;

[0026] According to the principle of principal component analysis, Decentralization, covariance matrix calculation and singular value decomposition are performed in sequence, and the point p is defined i (x i ,y i , z i ) is the eigenvector corresponding to the minimum eigenvalue, and the normal vector is specified The direction points to the positive direction of the coordinate axis;

[0027] 2.2 Rotate the surface A to be tested for surface flatness in the precast beam section to be parallel to the coordinate plane B:

[0028] A||B

[0029] The coordinate plane B is any plane among XOY, XOZ, and YOZ, and the reference vector is set is a unit vector perpendicular to the coordinate plane B. The angle between the normal vector and the reference vector at each point in the three-dimensional point cloud model Pt is calculated based on the vector angle formula. The i-th point p in the three-dimensional point cloud Pt i (x i ,y i , z i ) With the reference vector Angle θ i for:

[0030]

[0031] Set the angle threshold θ s is |θ s |≤5°, if the constraint conditions are met during the calculation: θ i ≤θ s , that is, θ i Not greater than the angle threshold θ s , then point p i (x i ,yi , z i ) is a point constituting the surface A to be measured, and this point is extracted into set A1. For the points in set A1, the coordinate values ​​of each point on the coordinate axis perpendicular to the surface A to be measured are taken to make a histogram, and the points contained in the interval with the largest frequency are extracted into set A2;

[0032] 2.3 According to the K nearest neighbor method, for the hth point p in set A2 h (x h ,y h , z h ), take the K nearest points around it to form a set Define point p h The point spacing d h For this point to the set Assuming the result is a Gaussian distribution, calculate the average distance between each point in set A2. And standard deviation σ:

[0033]

[0034] in, is the number of points in set A2, set is the distance threshold. If the constraints are met during the calculation process: That is, the point distance is not greater than the distance threshold, then the point p h (x h ,y h , z h ) are the points that make up the surface to be measured, and they are extracted into the surface to be measured A.

[0035] Preferably, the step (3) specifically comprises the following steps:

[0036] 3.1 Since the surface A to be measured is composed of three-dimensional coordinate points and has a certain thickness, a reference plane RP1 is made through the point on the surface A to be measured that is closest to the coordinate plane B and is parallel to the coordinate plane B. The reference plane RP1 is a1x+b1y+c1z+d1=0; a reference plane RP1 is made through the point on the surface A to be measured that is farthest from the coordinate plane B. n Parallel to coordinate plane B, reference plane RP n for a n x+b n y+c n z+d n =0; where a1, b1, c1, d1, a n , b n , c n , d n is the coefficient;

[0037] Set the iteration interval t to the point cloud density of the surface A to be measured, and move the reference plane from RP1 to RP in increments of t. n Translation, calculate the distance from each point in the measured surface A to the reference plane at each moving moment, the jth point p in the measured surface A j (x j ,y j , z j ) from the mth reference plane RP m The Euclidean distance r j for:

[0038]

[0039] Among them, N A is the number of midpoints of the surface A to be measured, N n is the number of reference planes, reference plane RP m The expression is a m x+b m y+c m z+d m =0, where a m , b m , c m , d m is the coefficient; according to the principle of least squares, the optimal reference plane RP is defined A is the r of each point on the surface A to be measured j The reference plane corresponding to the minimum sum is:

[0040]

[0041] Optimal reference plane RP A The expression is a A x+b A y+c A z+d A =0,a A , b A , c A , d A is the coefficient;

[0042] 3.2 For the jth point p in the surface A to be measured j (x j ,y j , z j ), the surface flatness at this point is f j Defined as point p j To the best reference plane RP A The Euclidean distance of:

[0043]

[0044] According to fj The numerical values ​​are marked with different colors to obtain a flatness distribution diagram of the surface A to be tested, thereby completing the surface flatness detection of the prefabricated beam section.

[0045] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0046] The present invention realizes fast and accurate detection of the flatness of prefabricated beam sections by performing coordinate calibration, target surface segmentation, and optimal reference surface determination on the three-dimensional point cloud model of the prefabricated beam section, which is beneficial to improving calculation efficiency and automation, avoiding the problems of traditional contact measurement that is prone to errors and low degree of result integration, saving calculation time, and improving the degree of digital integration of detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a method flow chart of a method for detecting surface flatness of a precast beam segment based on a three-dimensional point cloud model according to the present invention;

[0048] Figure 2 It is a schematic diagram of realizing precise calibration of coordinate axes according to slice features;

[0049] Figure 3 is the i-th point p i Normal vector n i Schematic diagram;

[0050] Figure 4 is a schematic diagram of the surface feature histogram;

[0051] Figure 5 It is a schematic diagram of the optimal reference plane search process. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0053] The present invention needs to meet the following constraints:

[0054] (1) The precast beam segments have regular geometric shapes. Generally speaking, the shapes of precast beam segments are regular basic shapes, such as rectangles and I-beams. The present invention uses a rectangular beam segment with a square cross-section as an example to illustrate the specific implementation method of the present invention.

[0055] (2) The placement position of the precast beam segment is basically horizontal, that is, the deviation between the original Z coordinate axis direction of the 3D point cloud model and the beam height direction is small. This constraint facilitates coordinate calibration.

[0056] (3) The origin of the three-dimensional rectangular coordinate system is set at the centroid of the three-dimensional point cloud model. The X-axis is parallel to the width of the beam, the Y-axis is parallel to the length of the beam, and the Z-axis is in the opposite direction of gravity and parallel to the height of the beam. For regular precast beam segment models, calibrating the three-dimensional rectangular coordinate system according to geometric features is more convenient for calculation and programming.

[0057] like Figure 1 As shown, the present invention provides a method for detecting the surface flatness of a prefabricated beam segment based on a three-dimensional point cloud model, which mainly includes the following steps:

[0058] (1) According to the specific form of the 3D point cloud model of the target component in the 3D rectangular coordinate system, the model is coarsely calibrated and finely calibrated in sequence, the spatial rotation matrix is ​​determined, and the point cloud coordinate calibration is performed;

[0059] (2) According to the principle of principal component analysis and K nearest neighbor method, the normal vector at each point of the component 3D point cloud model is determined, and the segmentation and extraction of a certain surface to be measured is realized by limiting the direction and coordinate range of the normal vector;

[0060] (3) According to the morphological relationship between the surface to be measured and the three-dimensional rectangular coordinate system, the optimal reference surface is iteratively found and the flatness of the surface is calculated.

[0061] In this embodiment, step 1 specifically includes the following steps:

[0062] 1.1 For the 3D point cloud model Pt0 of the precast beam segment, set the origin O of the original coordinate system X0Y0Z0 to the centroid position of the 3D point cloud model:

[0063]

[0064] Among them, Pt c is the 3D point cloud model after centralization; x1...x n is the coordinate value of each point X0 in Pt0, y1...y n is the Y0 coordinate value, z1...z n is the Z0 coordinate value; μ x is the average value of the X0 coordinates of each point in Pt0, μ Y is the average value of the original Y0 coordinate value, μ Z is the average value of the original Z0 coordinate value; the algorithm of the present invention is mainly aimed at prefabricated beam segments with regular geometric shapes. Setting the origin of the coordinate axis at the centroid position of the model can effectively simplify the calculation and improve efficiency.

[0065] 1.2 The algorithm of the present invention is mainly aimed at precast beam segments placed horizontally, so it can be approximately determined that the positive direction of the initial Z1 coordinate axis is the opposite direction of gravity and is approximately parallel to the beam height direction. cProject on the X0OY0 plane and perform principal component analysis on the projection. According to the principle of principal component analysis, firstly, the projected point cloud is decentralized, the covariance matrix of the decentralized point cloud is calculated, and the covariance matrix is ​​decomposed by singular value to obtain a set of eigenvalues ​​and eigenvectors uniquely corresponding to each eigenvalue. The eigenvector corresponding to the largest eigenvalue is the first principal component, and the eigenvector corresponding to the second largest eigenvalue is the second principal component. Define the Y1 coordinate axis direction as the first principal component direction and the X1 coordinate axis direction as the second principal component direction to complete the calibration of the initial coordinate system X1Y1Z1, that is, the rough calibration of the coordinates of the three-dimensional point cloud model.

[0066] 1.3 Reference Figure 2 , slices are made at appropriate positions that can reflect the characteristics of the beam width, beam length, and beam height of the component. The slice thickness is twice the point cloud density. The obtained slices should have the following characteristics: a. They have sufficient length, b. The point clouds contained in the slices should be as clear and continuous as possible; short slices or insufficient continuity can easily cause recognition errors during principal component analysis.

[0067] Project the point cloud contained in the slice onto the slice plane and perform principal component analysis. The projection can be approximated as a straight line. Compare the angles between the projection line and the two coordinate axes. According to the principle of principal component analysis, define the direction of the coordinate axis with the smaller angle with the projection line as the first principal component direction, and the direction of the other coordinate axis in the slice plane as the second principal component direction. After completing the above steps in the beam width, beam length, and beam height directions, the calibration of the final coordinate system XYZ is completed, that is, the precise calibration of the coordinates of the three-dimensional point cloud model.

[0068] According to the angle between the final coordinate system XYZ and the initial coordinate system X1Y1Z1, calculate the rotation matrix R corresponding to the change of the point cloud from the initial coordinate system X1Y1Z1 to the final coordinate system XYZ x , R y , R z If the rotation direction from the initial coordinate system X1Y1Z1 to the final coordinate system XYZ is counterclockwise, then the rotation matrix R is obtained. x , R y , R z They are:

[0069]

[0070] If the rotation direction is clockwise, the rotation matrix R x , R y , R z They are:

[0071]

[0072] Among them, α is the angle between coordinate system Y1OZ1 and coordinate system YOZ; β is the angle between coordinate system X1OZ1 and coordinate system XOZ, and γ is the angle between coordinate system X1OZ1 and coordinate system XOZ.

[0073] In summary, we get the spatial rotation matrix R = R x ·R y ·R z , and then realize the point cloud coordinate calibration:

[0074] Pt=R·Pt c

[0075] Among them, Pt c is the centered three-dimensional point cloud model, and Pt is the three-dimensional point cloud model after coordinate calibration.

[0076] In this embodiment, step 2 specifically includes the following steps:

[0077] 2.1 Reference Figure 3 According to the K nearest neighbor method, for the i-th point p in the 3D point cloud model Pt i (x i ,y i , z i ), select the nearest K neighboring points:

[0078]

[0079] Among them, N is the number of points in the 3D point cloud model Pt, is the set of K neighboring points closest to point pi; For neighboring points.

[0080] According to the principle of principal component analysis, Decentralization, covariance matrix calculation and singular value decomposition are performed in sequence, and the point p is defined i (x i ,y i , z i ) is the eigenvector corresponding to the minimum eigenvalue, and the normal vector is specified The direction points to the positive direction of the coordinate axis. The principle is that based on point p i (x i ,y i , z i )’s K neighboring points fit the plane and calculate the point p i The normal vector at is converted into the normal vector for calculating the above fitting plane.

[0081] 2.2 Rotate the surface A to be tested for surface flatness in the precast beam section to be parallel to the coordinate plane B:

[0082] A||B

[0083] The coordinate plane B is any plane among XOY, XOZ, and YOZ. Set the reference vector is a unit vector perpendicular to the coordinate plane B. The angle between the normal vector and the reference vector at each point in the three-dimensional point cloud model Pt is calculated based on the vector angle formula. The i-th point p in the three-dimensional point cloud Pt i (x i ,y i , z i ) Angle θ with reference vector i for:

[0084]

[0085] Set the angle threshold θ s is |θ s |≤5°. In the calculation process, if the constraint conditions are met: θ i ≤θ s , that is, θ i Not greater than the angle threshold θ s , then point p i (x i ,y i , z i ) is a point that constitutes the surface A to be measured, and this point is extracted to the set A1.

[0086] Reference Figure 4 For the points in set A1, take the coordinate values ​​of each point on the coordinate axis perpendicular to the surface A to be measured and make a histogram, and extract the points contained in the interval with the largest frequency to set A2.

[0087] 2.3 According to the K nearest neighbor method, for the hth point p in set A2 h (x h ,y h , z h ), take the K nearest points around it to form a set Define point p h The point spacing d h For this point to the set Assuming the result is a Gaussian distribution, calculate the average distance between each point in set A2. And standard deviation σ:

[0088]

[0089] in, is the number of points in set A2. Set is the distance threshold. If the constraints are met during the calculation process: That is, the point distance is not greater than the distance threshold, then the point ph (x h ,y h , z h ) are the points that make up the surface to be measured, and they are extracted into the surface to be measured A.

[0090] In this embodiment, step 3 specifically includes the following steps:

[0091] 3.1 Since the surface A to be measured is composed of three-dimensional coordinate points, it has a certain thickness. Through the point on the surface A to be measured that is closest to the coordinate plane B, make a reference plane RP1 (a1x+b1y+c1z+d1=0) parallel to the coordinate plane B; through the point on the surface A to be measured that is farthest from the coordinate plane B, make a reference plane RP n (a n x+b n y+c n z+d n =0) is parallel to coordinate plane B.

[0092] Reference Figure 5 , set the iteration interval t as the point cloud density of the surface A to be measured, and gradually move the reference plane from RP1 to RP with t as the increment n Translation, calculate the distance from each point in the measured surface A to the reference plane at each moving moment, the jth point p in the measured surface A j (x j ,y j , z j ) from the mth reference plane RP m (a m x+b m y+c m z+d m =0) j for:

[0093]

[0094] Among them, N A is the number of midpoints of the surface A to be measured, N n is the number of reference planes.

[0095] According to the least squares principle, the optimal reference plane RP is defined A (a A x+b A y+c A z+d A =0) are points r in the surface A to be measured j The reference plane corresponding to the minimum sum is:

[0096]

[0097] 3.2 For the jth point p in the surface A to be measured j (x j ,y j , z j ), the surface flatness at this point is f j Defined as point p j To the best reference plane RP A (a A x+b A y+c A z+d A =0) is:

[0098]

[0099] According to f j The numerical values ​​are marked with different colors, and the flatness distribution diagram of the surface A to be tested can be obtained, thereby completing the surface flatness detection of the prefabricated beam section.

[0100] This algorithm is based on multiple constraints and is therefore not suitable for surface flatness detection of all precast beam segments. However, for most commonly used types of precast beam segments in actual engineering, it can greatly improve the computational efficiency and automation of the flatness detection process compared to current traditional detection methods.

[0101] The technical solution of the present invention is further illustrated by using a specific case. The case calculation is mainly based on MATLAB to perform algorithm programming to obtain the calculation results.

[0102] The precast beam segment studied is a rectangular beam with a square cross-section, a beam length of 10m, a beam width of 2m, and a beam height of 2m. The point cloud density is 1 / cm 2 A 0.1m wide, 0.2m long area is preset on the top of the beam and moved 0.003m up from the original plane to simulate the unevenness of the prefabricated beam section. A rotation matrix is ​​pre-applied to it:

[0103]

[0104]

[0105] 1. Centralize the 3D point cloud model of the precast beam segment and perform rough calibration based on its initial shape to obtain the rotation matrix Then, according to step 1.3, the three-dimensional point cloud model is precisely calibrated in the three coordinate axis directions in turn to obtain the rotation matrix:

[0106]

[0107]

[0108] The same as the preset rotation matrix.

[0109] 2. According to step 2.1, calculate the normal vector at each point. To extract the point cloud contained in the top surface, set the reference normal vector Calculate the angle between the normal vector at each point and the reference normal vector. Set the angle threshold θ s ±5°, and judge whether the normal vector at each point meets the standard and satisfies the judgment condition θ i ≤θ s Extract the points to set A2; take the z coordinates of the points in set A2 and make a histogram with a width of 0.1m, and extract the points corresponding to the cells with more than 5000 3D points and z coordinates greater than 0. According to step 2.3, calculate the distance, average distance and standard deviation of each extracted point, and filter out those points exceeding the distance threshold. The final top surface point cloud is obtained.

[0110] 3. According to step 3.1, make a reference plane z = minz and move it towards z = maxz in increments of 0.0001m. For each reference plane z = z i , calculate the distance from each point to the reference plane. Finally, the optimal reference plane is determined to be z=0.9991, and the distance between the uneven part and the optimal reference plane is 0.003m, which is consistent with the initial setting.

[0111] The symbols and parameters involved in the present invention are listed as follows:

[0112]

[0113] The above content is only a further detailed description of the present invention through a specific embodiment, and the present invention cannot be considered to be limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for detecting the surface flatness of precast beam segments based on a three-dimensional point cloud model. Features: The following steps are involved: (1) According to the specific form of the 3D point cloud model of the target component in the 3D rectangular coordinate system, the model is coarsely calibrated and finely calibrated in sequence, the spatial rotation matrix is ​​determined, and the point cloud coordinate calibration is performed; (2) According to the principle of principal component analysis and K nearest neighbor method, the normal vector at each point of the component 3D point cloud model is determined, and the segmentation and extraction of a certain surface to be measured is realized by limiting the direction and coordinate range of the normal vector; (3) Iteratively searching for the optimal reference surface and calculating the flatness of the surface based on the morphological relationship between the surface to be measured and the three-dimensional rectangular coordinate system; Step (3) specifically includes the following steps: 3.1 Since the surface A to be measured is composed of three-dimensional coordinate points and has a certain thickness, the reference plane RP is made through the point on the surface A to be measured that is closest to the coordinate plane B. 1 Parallel to coordinate plane B, reference plane RP 1 for a 1 x+b 1 y+c 1 z+d 1 = 0; make a reference plane RP through the point on the measured surface A that is farthest from the coordinate plane B n Parallel to coordinate plane B, reference plane RP n for a n x+b n y+c n z+d n =0; where a 1 , b 1 , c 1 , d 1 , a n , b n , c n , d n is the coefficient; Set the iteration interval t to the point cloud density of the surface A to be measured, and change the reference plane from RP 1 Step by step, move towards RP n Translation, calculate the distance from each point in the measured surface A to the reference plane at each moving moment, the jth point p in the measured surface A j (x j ,y j ,z j ) from the mth reference plane RP m The Euclidean distance r j for: Among them, N A is the number of midpoints of the surface A to be measured, N n is the number of reference planes, reference plane RP m The expression is a m x+b m y+c m z+d m =0, where a m , b m , c m , d m is the coefficient; according to the principle of least squares, the optimal reference plane RP is defined A is the r of each point on the surface A to be measured j The reference plane corresponding to the minimum sum is: Optimal reference plane RP A The expression is a A x+b A y+c A z+d A =0,a A , b A , c A , d A is the coefficient; 3.2 For the jth point p in the surface A to be measured j (x j ,y j ,z j ), the surface flatness at this point is f j Defined as point p j To the best reference plane RP A The Euclidean distance of: According to f j The numerical values ​​are marked with different colors to obtain a flatness distribution diagram of the surface A to be tested, thereby completing the surface flatness detection of the prefabricated beam section.

2. According to the method for detecting the surface flatness of a precast beam segment based on a three-dimensional point cloud model according to claim 1, It is characterized in that Step (1), step (2), and step (3) must all satisfy the following constraints: the geometric shape of the precast beam segment is regular, the placement position of the precast beam segment is horizontal, the origin of the three-dimensional coordinate system is set at the centroid of the point cloud model, and the X-axis, Y-axis, and Z-axis directions of the three-dimensional coordinate system are parallel to the component beam width, beam length, and beam height directions, respectively.

3. The method for detecting surface flatness of a precast beam segment based on a three-dimensional point cloud model according to claim 1, It is characterized in that Step (1) specifically includes the following steps: 1.1 Three-dimensional point cloud model of precast beam segment Pt 0 , set the original coordinate system X 0 Y 0 Z 0 The origin O is at the centroid of the 3D point cloud model: Among them, Pt c is the 3D point cloud model after centralization; x 1 …x n Pt 0 Each point X 0 Coordinate value, y 1 …y n Y 0 Coordinate value, z 1 …z n Z 0 Coordinate value; μ X Pt 0 Each point X 0 The average value of the coordinates, μ Y is the original Y 0 The average value of the coordinates, μ z The original Z 0 The average value of the coordinates; 1.2 For the horizontally placed precast beam segment, determine the initial Z 1 The positive direction of the coordinate axis is the opposite direction of gravity and is parallel to the beam height direction. c In X 0 OY 0 Plane projection, perform principal component analysis on the projection. According to the principle of principal component analysis, firstly, the projected point cloud is decentralized, the covariance matrix of the decentralized point cloud is calculated, and the covariance matrix is ​​decomposed by singular value to obtain a set of eigenvalues ​​and the eigenvectors uniquely corresponding to each eigenvalue. The eigenvector corresponding to the largest eigenvalue is the first principal component, and the eigenvector corresponding to the second largest eigenvalue is the second principal component. Define Y 1 The coordinate axis direction is the direction of the first principal component, X 1 The direction of the coordinate axis is the direction of the second principal component, completing the initial coordinate system X 1 Y 1 Z 1 Calibration, that is, rough calibration of coordinates of the 3D point cloud model; 1.3 Slices are made at appropriate positions that can reflect the characteristics of beam width, beam length, and beam height of the component. The slice thickness is twice the point cloud density. The point cloud contained in the slice is projected onto the slice plane and principal component analysis is performed. The projection is fitted into a straight line. The angles between the projection line and the two coordinate axes are compared. According to the principle of principal component analysis, the direction of the coordinate axis with the smallest angle with the projection line is defined as the first principal component direction, and the direction of the other coordinate axis in the slice plane is defined as the second principal component direction. After completing the above steps in the beam width, beam length, and beam height directions, the calibration of the final coordinate system XYZ is completed, that is, the precise calibration of the coordinates of the three-dimensional point cloud model. According to the final coordinate system XYZ and the initial coordinate system X 1 Y 1 Z 1 The angle between the points is calculated from the initial coordinate system X 1 Y 1 Z 1 Change to the rotation matrix R corresponding to the final coordinate system XYZ x ,R y ,R z , if from the initial coordinate system X 1 Y 1 Z 1 The rotation direction of the final coordinate system XYZ is counterclockwise, and the rotation matrix R is obtained. x ,R y ,R z They are: If the rotation direction is clockwise, the rotation matrix R x ,R y ,R z They are: Among them, α is the coordinate system Y 1 OZ 1 The angle with coordinate system YOZ; β is the angle with coordinate system X 1 OZ 1 The angle with coordinate system XOZ, γ is the angle of coordinate system X 1 OZ 1 Angle with coordinate system XOZ; According to the above R x ,R y ,R z Get the spatial rotation matrix R = R x ·R y ·R z , and then realize the point cloud coordinate calibration: Pt=R·Pt c Among them, Pt c is the centered three-dimensional point cloud model, and Pt is the three-dimensional point cloud model after coordinate calibration.

4. The method for detecting surface flatness of a precast beam segment based on a three-dimensional point cloud model according to claim 3, It is characterized in that Step (2) The following steps are involved: 2.1 According to the K nearest neighbor method, for the i-th point p in the 3D point cloud model Pt i (x i ,y i ,z i ), select the nearest K neighboring points: in, N is the number of points in the 3D point cloud model Pt, is the distance from point p i The set of the nearest K neighboring points, is the kth neighboring point; According to the principle of principal component analysis, Decentralization, covariance matrix calculation and singular value decomposition are performed in sequence, and the point p is defined i (x i ,y i ,z i ) is the eigenvector corresponding to the minimum eigenvalue, and the normal vector is specified The direction points to the positive direction of the coordinate axis; 2.2 Rotate the surface A to be tested for surface flatness in the precast beam section to be parallel to the coordinate plane B: A∥B The coordinate plane B is any plane among XOY, XOZ, and YOZ, and the reference vector is set is a unit vector perpendicular to the coordinate plane B. The angle between the normal vector and the reference vector at each point in the three-dimensional point cloud model Pt is calculated based on the vector angle formula. The i-th point p in the three-dimensional point cloud Pt i (x i ,y i ,z i ) With the reference vector Angle θ i for: Set the angle threshold θ s is |θ s |≤5°, if the constraint conditions are met during the calculation: θ i ≤θ s , that is, θ i Not greater than the angle threshold θ s , then point p i (x i ,y i ,z i ) is the point that constitutes the surface A to be measured, and this point is extracted to the set A 1 , for the set A 1 Take the coordinate values ​​of each point on the coordinate axis perpendicular to the surface A to be measured and make a histogram, extract the points contained in the interval with the largest frequency to set A 2 ; 2.3 According to the K nearest neighbor method principle, for set A 2 The hth point p h (x h ,y h ,z h ), take the K nearest points around it to form a set Define point p h The point spacing d h For this point to the set The average value of the distance between each point in the set A is assumed to be Gaussian distributed. 2 The average distance between each point And standard deviation σ: in, For set A 2 The number of midpoints, set is the distance threshold. If the constraints are met during the calculation process: That is, the point distance is not greater than the distance threshold, then the point p h (x h ,y h ,z h ) are the points that make up the surface to be measured, and they are extracted into the surface to be measured A.

Citation Information

Cited By

  • High-speed rail train body surface flatness detection method and system

    CN122023266A