A method for synchronous extraction of reinforcement cross-welding points and welding seam features based on point clouds

Abstract

The application discloses a method for synchronously extracting steel bar cross welding points and welding seam features based on point clouds, and relates to the technical field of welding detection. The method comprises the following steps: acquiring steel bar point clouds of a steel bar to be detected based on a 3D scanner; and performing pretreatment to obtain point clouds without discrete points; performing Euclidean clustering on the point clouds without discrete points based on the maximum and minimum values of the number of clustered points, so as to obtain a plurality of complete steel bar point sets and a plurality of incomplete steel bar point sets; performing merging processing on the plurality of incomplete steel bar point sets, so as to obtain corresponding complete steel bar point sets; screening complete steel bar point sets containing welding seam features from all the complete steel bar point sets; performing segmentation on any complete steel bar point set containing welding seam features, so as to obtain corresponding steel bar point sets and remaining steel bar point sets; obtaining welding seam features and final steel bar point sets of the steel bar to be detected according to each pair of steel bar point sets and remaining steel bar point sets; and obtaining welding point space coordinates according to each two final steel bar point sets. The application realizes synchronous extraction of steel bar cross welding points and welding seam features.

Description

Technical Field

[0001] This invention relates to the field of welding inspection technology, and in particular to a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point cloud. Background Technology

[0002] With the development of the social economy, automation and robotics technologies are also rapidly advancing, and welding robots are being widely used in welding environments to replace manual labor. Reinforcing bars in buildings are not only arranged in crisscrossing patterns but also include a large number of parallel reinforcing bar structures. During the welding process, the main identification methods used by robots include traditional geometry-based identification of intersecting weld points and image-based deep learning methods for identifying weld features.

[0003] Current methods for extracting weld features of reinforcing bars mainly involve scanning point cloud data provided by a 3D scanner, calculating its spatial geometric information, and processing it to obtain the desired cross-weld point features. Image-based deep learning methods can effectively identify cross-weld points and weld seam features.

[0004] The main drawbacks of existing technologies are:

[0005] 1) It has strong specificity. Currently, the traditional geometric-based cross weld point feature extraction is only applicable to the extraction of cross weld point features, which has strong specificity, but cannot extract weld seam features at the same time.

[0006] 2) High computational load. The classification and merging of point sets after Euclidean clustering of the reinforcement point cloud involves a large number of calculations of cylindrical parameters and axial relationships, which increases the computational load.

[0007] 3) Weld feature extraction is relatively difficult. Traditional Euclidean clustering cannot segment point clouds of side-by-side rebars with weld features, making it impossible to extract these features. To date, there is limited research on methods for weld feature extraction.

[0008] 4) Image-based deep learning methods suffer from instability and a failure to achieve 100% recognition rate. Furthermore, due to the lack of depth information in image recognition and the susceptibility to background influence during pixel-level segmentation, the accuracy of identified welding features is not high. Summary of the Invention

[0009] The purpose of this invention is to provide a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point cloud, thereby achieving simultaneous extraction of cross weld points and weld features of reinforcing bars.

[0010] To achieve the above objectives, the present invention provides the following solution:

[0011] A method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point clouds, the simultaneous extraction method comprising:

[0012] Rebar point cloud acquisition based on 3D scanner;

[0013] The point cloud of the reinforcing bars is preprocessed to obtain a point cloud without discrete points;

[0014] Based on the maximum and minimum values ​​of the number of cluster points, Euclidean clustering is performed on point clouds without discrete points, resulting in multiple complete rebar point sets and multiple incomplete rebar point sets.

[0015] Multiple incomplete sets of rebar points are merged to obtain the corresponding complete set of rebar points.

[0016] Select the complete set of rebar points containing weld features from all complete rebar point sets;

[0017] Any complete set of reinforcing bars containing weld features is segmented to obtain the corresponding set of reinforcing bars and the remaining set of reinforcing bars;

[0018] Based on each pair of rebar point sets and the remaining rebar point sets, the weld characteristics of the rebar to be tested and the final rebar point set are obtained;

[0019] The spatial coordinates of the weld point are obtained from every two final sets of reinforcing steel points.

[0020] Optionally, the point cloud of the reinforcing bars is preprocessed to obtain a point cloud without discrete points, specifically including:

[0021] The point cloud of the reinforcing bars is subjected to pass-through filtering and point cloud plane segmentation to obtain the point cloud of welding feature points;

[0022] By removing outliers from the point cloud of welding feature points using statistical analysis techniques, a point cloud without discrete points is obtained.

[0023] Optionally, the step of merging multiple incomplete sets of reinforcing bars to obtain a corresponding complete set of reinforcing bars specifically includes:

[0024] Multiple structures are constructed based on multiple incomplete sets of rebar points using PCA principal component analysis. Each structure includes an incomplete set of rebar points and endpoint information of the rebar points. Each set of rebar points includes two endpoint information and axis information.

[0025] A complete set of reinforcement points is obtained by classifying multiple structures.

[0026] Optionally, the process of classifying multiple structures to obtain a complete set of reinforcement points specifically includes:

[0027] For each structure, classify it against the remaining structures according to the classification criteria; merge the structures that meet the criteria into a complete set of reinforcement points; the structures that do not meet the criteria are the remaining structures.

[0028] Each of the remaining structures is compared with the other structures according to the classification criteria; the structures that meet the classification criteria are merged into another complete set of reinforcement points.

[0029] Repeat the process until the remaining structures have completed the category condition judgment or all structures have been merged.

[0030] Optionally, the category conditions are specifically:

[0031] The distance between the nearest two endpoints of any structure and any other structure is less than the endpoint distance threshold;

[0032] The angle between any structure and the axis vector of any other structure is less than the angle threshold.

[0033] Optionally, any complete set of reinforcing bars containing weld features can be divided into a corresponding set of reinforcing bars and a set of remaining reinforcing bars, specifically including:

[0034] A pass-through filter is applied to the complete set of rebar points containing weld features to obtain a complete set of rebar points without other feature information.

[0035] For any complete set of rebar points without other feature information, the vector and radius of the rebar centerline are obtained by the least squares method.

[0036] For any complete set of rebar points without other feature information, determine whether the distance from the point to the axis is less than the point-to-line distance threshold; if not, remove the point; if yes, obtain the set of rebar points.

[0037] The remaining set of reinforcement points is obtained from any set of reinforcement points using an octree.

[0038] Optionally, obtaining the weld characteristics of the reinforcing bar to be tested based on each pair of reinforcing bar point sets and the remaining reinforcing bar point sets specifically includes:

[0039] Based on the PCA algorithm, four points are obtained from each pair of rebar point sets and the remaining rebar point sets; the four points are the two farthest points in the rebar point set and the two farthest points in the remaining rebar point set.

[0040] The projected coordinates of the four points are obtained using the least squares method;

[0041] Project the coordinates of the four points onto any central axis, and obtain the weld features of the weld area based on the two points where the weld area appears.

[0042] Optionally, obtaining the final set of reinforcing bars based on each pair of reinforcing bar point sets and the remaining set of reinforcing bar point sets specifically includes:

[0043] Based on the PCA algorithm, four points are obtained from each pair of rebar point sets and the remaining rebar point sets; the four points are the two farthest points in the rebar point set and the two farthest points in the remaining rebar point set; the final rebar point set includes four points.

[0044] Optionally, obtaining the spatial coordinates of the weld point based on every two final reinforcement point sets specifically includes:

[0045] Cylindrical fitting is performed on each final set of reinforcement points to obtain the axis, radius, and a point on the axis;

[0046] The spatial coordinates of the weld point are obtained by comparing a point on the axis of any final set of reinforcement points with another point on the axis of the nearest final set of reinforcement points.

[0047] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0048] This invention discloses a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point clouds. The method includes: acquiring a point cloud of the reinforcing bar to be tested using a 3D scanner; preprocessing it to obtain a point cloud without discrete points; performing Euclidean clustering on the point cloud without discrete points based on the maximum and minimum values ​​of the clustering points, resulting in multiple complete and multiple incomplete point sets of reinforcing bars; merging the multiple incomplete point sets to obtain corresponding complete point sets of reinforcing bars; selecting complete point sets containing weld features from all complete point sets; segmenting any complete point set containing weld features to obtain corresponding point sets and remaining point sets of reinforcing bars; obtaining the weld features and final point sets of the reinforcing bar to be tested based on each pair of point sets and remaining point sets; and obtaining the spatial coordinates of the weld points based on every two final point sets of reinforcing bars. Compared with existing technologies that are only applicable to the extraction of cross weld point features and cannot simultaneously extract weld features, this invention achieves simultaneous extraction of cross weld points and weld features of reinforcing bars.

[0049] Compared with existing technologies that rely on image deep learning to identify welding features with low accuracy, this invention ensures the accuracy of coordinates by providing point cloud data through a 3D scanner. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a schematic diagram of the method flow for the synchronous extraction of cross weld points and weld features of reinforcing bars based on point cloud according to the present invention;

[0052] Figure 2 This is a point-end projection schematic diagram of a method for synchronously extracting cross weld points and weld features of reinforcing bars based on point clouds according to the present invention.

[0053] Figure 3 This is a schematic diagram illustrating the weld feature solving method of the present invention, which is a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point cloud.

[0054] Figure 4 This is a welding feature structure diagram of a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point cloud according to the present invention.

[0055] Figure 4 (a) is a diagram of the first welding feature structure;

[0056] Figure 4 (b) is a diagram of the second welding feature structure;

[0057] Figure 4 (c) is a diagram of the third welding feature structure;

[0058] Figure 5 This is a schematic diagram of step S4 of the present invention, which is a method for synchronously extracting cross weld points and weld features of reinforcing bars based on point clouds. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The purpose of this invention is to provide a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point cloud, thereby achieving simultaneous extraction of cross weld points and weld features of reinforcing bars.

[0061] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] like Figure 1 As shown, this invention provides a method for simultaneous extraction of cross weld points and weld features of reinforcing bars based on point clouds. The simultaneous extraction method includes:

[0063] S1: Obtain the point cloud of the reinforcing bar to be tested based on a 3D scanner.

[0064] S2: Preprocess the point cloud of the reinforcing bars to obtain a point cloud without discrete points.

[0065] Step S2 specifically includes:

[0066] S21: Perform pass-through filtering and point cloud plane segmentation on the point cloud of the reinforcing bars to obtain the point cloud of welding feature points.

[0067] S22: Based on statistical analysis techniques, outliers are removed from the point cloud of the welding feature points to obtain a point cloud without discrete points.

[0068] S3: Based on the maximum and minimum values ​​of the number of cluster points, perform Euclidean clustering on point clouds without discrete points, resulting in multiple complete rebar point sets and multiple incomplete rebar point sets.

[0069] S4: Merge multiple incomplete rebar point sets to obtain the corresponding complete rebar point set.

[0070] Step S4 specifically includes:

[0071] S41: Construct multiple structures based on multiple incomplete rebar point sets using PCA principal component analysis; each structure includes an incomplete rebar point set and rebar point set endpoint information; each rebar point set includes two endpoint information and axis information.

[0072] S42: Classify multiple structures to obtain a complete set of reinforcement points.

[0073] Step S42 specifically includes:

[0074] S421: For any given structure, classify it against the remaining structures according to the classification criteria; merge the structures that meet the criteria into a complete set of reinforcement points; the structures that do not meet the criteria are the remaining structures.

[0075] S422: Distinguish between any remaining structure and the remaining structures according to their classification conditions; merge the structures that meet the classification conditions into another complete set of reinforcement points.

[0076] S423: Repeat until all complete reinforcement point sets are obtained, where any complete reinforcement point set includes at least two structures.

[0077] Specifically, the category conditions are as follows:

[0078] The distance between the nearest two endpoints of any structure and any other structure is less than the endpoint distance threshold.

[0079] The angle between any structure and the axis vector of any other structure is less than the angle threshold.

[0080] S5: Select the complete set of rebar points containing weld features from all complete rebar point sets.

[0081] S6: Segment any complete set of reinforcing bars containing weld features to obtain the corresponding set of reinforcing bars and the remaining set of reinforcing bars.

[0082] Step S6 specifically includes:

[0083] S61: Perform pass-through filtering on the complete set of rebar points containing weld features to obtain a complete set of rebar points without other feature information.

[0084] S62: For any complete set of rebar points without other feature information, obtain the rebar centerline vector and radius using the least squares method.

[0085] S63: For any complete set of rebar points without other feature information, determine whether the distance from the point to the axis is less than the point-to-line distance threshold; if not, remove the point; if yes, obtain the set of rebar points.

[0086] S64: Obtain the remaining set of rebar points based on any set of rebar points using an octree.

[0087] S7: Based on each pair of rebar point sets and the remaining rebar point sets, obtain the weld characteristics of the rebar to be tested and the final rebar point set.

[0088] Step S7 specifically includes:

[0089] S71: Based on the PCA algorithm, four points are obtained from each pair of rebar point sets and the remaining rebar point sets; the four points are the two farthest points in the rebar point set and the two farthest points in the remaining rebar point set. The final rebar point set includes four points.

[0090] S72: Obtain the projected coordinates of the four points based on the least squares method.

[0091] S73: Project the projection coordinates of the four points onto any central axis, and obtain the weld features of the weld area based on the two points where the weld area appears.

[0092] S8: Obtain the spatial coordinates of the weld point based on every two final reinforcement point sets.

[0093] Step S8 specifically includes:

[0094] S81: Perform cylindrical fitting on each final set of reinforcement points to obtain the axis, radius, and a point on the axis.

[0095] S82: Obtain the spatial coordinates of the weld point based on a point on the axis of any final reinforcement set and another point on the axis of the nearest final reinforcement set.

[0096] This invention utilizes a KD-tree-like search mechanism to merge rebar point sets in the same direction using Euclidean clustering. By leveraging the geometric information of the cluster endpoints and the axis direction of the point set, it quickly searches for and classifies the rebar point sets in the same direction, thereby achieving rapid classification and merging of rebar point clouds in the same direction.

[0097] The following specific embodiments are examples:

[0098] like Figure 4 As shown, the welding features of the present invention include: Figure 4 As shown in (a), 1. Characteristics of cross solder joints; as... Figure 4 (b) shows the weld characteristics; as shown in Figure 2. Figure 4 As shown in (c), 3. Features of shielded weld.

[0099] The point cloud of the reinforcing steel bar to be tested is obtained using a 3D scanner.

[0100] The rebar point cloud is preprocessed by using a pass-through filter to remove points without welding features. Then, background point cloud is removed through point cloud plane segmentation to obtain a point cloud with welding features. Finally, a Statistical Outlier Removal filter is used to remove outliers, resulting in a point cloud without discrete points. Specifically, statistical analysis techniques are used to perform a statistical analysis on the neighborhood of each point and prune points that do not meet certain criteria. The sparse outlier removal method is based on calculating the distance distribution from a point to its neighbors in the input data. For each point, the average distance to all its neighbors is calculated. Assuming the result follows a Gaussian distribution, the shape of which is determined by the mean and standard deviation, points whose average distance is outside the standard range are defined as outliers and removed from the data.

[0101] Based on the maximum and minimum values ​​of the number of cluster points, Euclidean clustering is performed on point clouds without discrete points, resulting in multiple complete sets of rebar points and multiple incomplete sets of rebar points P1, P2, P3…P… n , where n is the number of incomplete reinforcement point sets.

[0102] Multiple incomplete reinforcement point sets are merged. For example... Figure 5 As shown, multiple structures are constructed based on multiple incomplete rebar point sets using Principal Component Analysis (PCA). Each structure includes an incomplete rebar point set and endpoint information for each set. Each rebar point set includes two endpoints and axis information. By analyzing the geometric relationship between each incomplete rebar point set and its surrounding incomplete rebar point sets, the two closest endpoints in the two rebar point sets are found through a query. Simultaneously, the distance between these two endpoints is less than an endpoint distance threshold, and the angle between the axis vectors is less than an angle threshold. The structures containing the two rebar point sets are then classified as belonging to the same category. Subsequently, based on the above conditions, structures with all approximately unidirectional rebar point sets are merged into a single complete rebar point set. Finally, multiple complete rebar point sets Q1, Q2, Q3…Q are obtained. r , where r is the number of complete reinforcement point sets.

[0103] For any complete set of rebar points containing weld features, segmentation is performed to obtain the corresponding rebar point set and the remaining rebar point set. First, a pass-through filter is used to segment and filter the complete set of rebar points containing weld features to obtain the complete set of rebar points Q without other feature information. i Let 1 < i ≤ r, where i is the i-th complete set of rebar points without other feature information; the cylinder fitting method is used to calculate the central axis and radius of the rebar. According to the least squares principle, the sum of the squares of the distances from each point to the fitted line is minimized; the specific formula is: k represents the set of reinforcement points Q i Number of midpoints; d s Let l represent the distance from the s-th point to the fitted line; where l 2 +m 2 +n 2 =1; l, m, n represent the three components of the direction vector of the fitted line.

[0104] The axis passes through the centroid of each point, so we can solve for:

[0105]

[0106] Where δ x δ y δ z These are the increments in the x, y, and z directions, respectively, representing the centroid.

[0107] After equation transformation, we obtain:

[0108] f(l,m,n)=(1-l 2 B 11 +(1-m 2 B 22 +(1-n 2 B 33 -2lmB 12 -2nlB 13 -2mnB 23 .

[0109] in, The transformation involves finding the minimum value of f(l,m,n). When (l,m,n) is the unit eigenvector corresponding to the largest eigenvalue λmax(B) of matrix B, f(l,m,n) can take its minimum value. This unit eigenvector is the unit direction vector of the cylinder axis; where λ represents the eigenvalue of matrix B.

[0110] Let there be multiple complete sets of reinforcement points Q1, Q2, Q3…Q r Among them, the i-th point (x i ,y i ,z iLet the coordinate system of point Q be {O-XYZ}, and let i be any point in the complete set of reinforcing bars. Perform a rotation transformation on point Q, and let the coordinate system after the transformation be {O-X'Y'Z'}, so that the Z' axis coincides with the vector (l,m,n). The rotation transformation matrix R is:

[0111]

[0112] Solve for the coordinates of the transformed point set: NEWxyz=R×(x i ,y i ,z i ).

[0113] The cylinder is now perpendicular to the {O-XY} plane, and its projection onto the {O-XY} plane is a circle. Therefore, we can consider only the x and y coordinates and use the least squares method to solve for the radius and center coordinates of this circle. Assume the least squares fitted circular curve is r. 2 =(yb) 2 +(xa) 2 The coordinates of the center of the circle are (a, b) and the radius is r. Based on the coordinates after coordinate transformation, the radius r and center (a, b) of the projected circle are calculated. The center z = 0 in the coordinate system {O-X'Y'Z'}. Then, an inverse coordinate transformation is performed on the center coordinates (a, b, 0) in the coordinate system {O-X'Y'Z'} to obtain a point q on the axis of the cylinder.

[0114] Based on the axis vector and radius of the reinforcement point set obtained above, a distance threshold between the point and the axis is set. According to the constraint that the distance from the point to the axis is less than the threshold (Distance < R), from the point set Q... i The first set of reinforcement points Q is calculated and extracted. c1 Then, the spatial change detection method for unorganized point cloud data using Octree Point Cloud Change Detector (octree is a tree structure for managing sparse 3D data. octree is used to detect spatial changes between multiple unordered point clouds, which may differ in size, resolution, density, and point order. By recursively comparing the tree structure of octree, the spatial changes represented by the differences in voxel composition generated by octree can be identified) is used to quickly extract the rebar point set Q. c1 The point set Q that is not present in the constructed octree c2 That is, the remaining set of steel reinforcement points Q c2 , the point set Q c1 Q c2 Output.

[0115] Based on each pair of rebar point sets and the remaining rebar point sets, the weld characteristics of the rebar to be tested and the final rebar point set are obtained. Specifically: based on two rebar point sets Q with weld characteristics... c1 Q c2 The spatial positions of the central axis endpoints are used to extract the two endpoints of the corresponding weld. First, the point set Q is calculated using the PCA algorithm. c1 Two farthest points q1 and q2 are used, and then the point set Q is calculated using the previous least squares method. c1 Given an axis α with vectors (l, m, n) and a point q with coordinates (a, b, c) on the axis, solve for the parameter t = α·(q-q1) / α. 2 ;like Figure 2 As shown, the coordinates of q'1, projected from q1, are (lt+a, mt+b, nt+c), and the coordinates of q'2 are obtained similarly. Q can be calculated similarly. c2 The coordinates of q'3 and q'4 are calculated. Finally, q1', q'2, q'3, and q'4 are projected onto any central axis; for example... Figure 3 As shown, by solving for the two points p1 and p2 in the weld region, the weld characteristics of the weld region can be calculated. Here, p1 and p2 are the weld characteristic lines. The two endpoints ( and (Not parallel).

[0116] Finally, complete the set of all complete rebar points Q. i The weld features are extracted and segmented to obtain the weld features and the final set of reinforcement points T1, T2, T3…T e , where e is the number of the final reinforcement point set.

[0117] For each final set of reinforcement points, a cylindrical fit is performed to obtain the axis, radius, and a single point on the axis. The spatial coordinates of the weld point are obtained based on the axis point on any given final set of reinforcement points and the axis point on the nearest other final set of reinforcement points. The specific calculation formula is as follows: Where (p2-p1) v32 The vector representing the points on the point cloud axis of two final reinforcement point sets in v 32 The projection of the vector onto v 32 The axis is perpendicular to v2 and the axes of the two reinforcing bars v1×v2; similarly, v 31 The axis is perpendicular to v1 and the two reinforcing bar axes v1×v2. t1 and t2 are the two points closest to the two axes; (t1+t2) / 2 is the desired cross weld point.

[0118] To verify the accuracy of the cross-welding points, assume that the set of rebar points T calculated by t3 is... a T b The welding intersection must be at the same time at the rebar point set T.a T b On the central axis, the point set T a T b The endpoints of the central axis are t a1 , t a2 , t b1 , t b2 Calculate (t3-t) respectively a1 ) 2 +(t3-t a2 ) 2 ≤(t a1 -t a2 ) 2 Similarly, (t3-t) b1 ) 2 +(t3-t b2 ) 2 ≤(t b1 -t b2 ) 2 If both conditions are met, the correct position of the required welding intersection point t3 can be determined.

[0119] In the process of extracting features of parallel weld seams of reinforcing bars, this invention divides one set of reinforcing bar points according to a radius threshold method, and divides the remaining set of reinforcing bar points using an octree voxel mesh to remove background points. The weld seam features between the parallel reinforcing bars are calculated geometrically to achieve rapid extraction of the weld seam starting coordinates.

[0120] This invention calculates cross weld points and weld features based on traditional geometry of three-dimensional point clouds. It can quickly extract single cross weld features, single weld features, composite cross weld points, general weld features, and weld features partially obscured by reinforcing bars in various situations.

[0121] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0122] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for simultaneously extracting reinforcement cross-welding points and welding seam features based on a point cloud, characterized in that, The synchronous extraction method comprises: Obtaining a steel bar point cloud of a to-be-tested steel bar based on a 3D scanner; Pretreating the steel bar point cloud to obtain a non-discrete point cloud; Based on the maximum and minimum values of the number of clustered points, performing Euclidean clustering on the non-discrete point cloud to obtain a plurality of complete steel bar point sets and a plurality of incomplete steel bar point sets; Merging the plurality of incomplete steel bar point sets to obtain a corresponding complete steel bar point set, specifically comprising: Based on a PCA principal component analysis method, constructing a plurality of structures from the plurality of incomplete steel bar point sets; each structure comprises an incomplete steel bar point set and steel bar point set endpoint information; each steel bar point set comprises two endpoint information and axis information; Classifying the plurality of structures to obtain a complete steel bar point set, specifically comprising: Respectively performing category condition discrimination on any structure and the remaining structures; merging the structures meeting the discrimination condition into one complete steel bar point set; the structures not meeting the discrimination condition are the remaining structures; Respectively performing category condition discrimination on any structure in the remaining structures and the remaining structures; merging the structures meeting the discrimination condition into another complete steel bar point set; Repeating the execution until the remaining structures perform the category condition discrimination or all the structures are merged; The category condition is specifically: The distance between the two nearest endpoints of any structure and the remaining structures is less than the endpoint distance threshold value; The angle between the axis vectors of any structure and the remaining structures is less than the angle threshold value; Screening a complete steel bar point set containing a weld seam feature from all the complete steel bar point sets; Segmenting any complete steel bar point set containing a weld seam feature to obtain a corresponding steel bar point set and a remaining steel bar point set; Obtaining a weld seam feature and a final steel bar point set of the to-be-tested steel bar according to each pair of steel bar point set and remaining steel bar point set; Obtaining a weld point space coordinate according to each two final steel bar point sets.

2. The method for point cloud based reinforcement cross-weld and weld feature synchronous extraction according to claim 1, characterized in that, The pretreatment of the steel bar point cloud to obtain a non-discrete point cloud specifically comprises: Performing straight-through filtering and point cloud plane segmentation on the steel bar point cloud to obtain a welding feature point cloud; Based on a statistical analysis technique, removing outliers from the welding feature point cloud to obtain a non-discrete point cloud.

3. The method for point cloud based reinforcement cross-weld and weld feature synchronous extraction according to claim 1, characterized in that, Segmenting any complete steel bar point set containing a weld seam feature into a corresponding steel bar point set and a remaining steel bar point set, specifically comprising: Performing straight-through filtering on the complete steel bar point set containing a weld seam feature to obtain a complete steel bar point set without other feature information; For any complete steel bar point set without other feature information, obtaining a steel center axis vector and a radius according to a least square method; For any complete steel bar point set without other feature information, respectively judging whether the distance of a point to the axis is less than a point-line distance threshold value; if not, the point is removed; if yes, a steel bar point set is obtained; Based on any steel bar point set, obtaining a remaining steel bar point set according to an octree.

4. The method for point cloud based reinforcement cross-weld and weld feature synchronous extraction according to claim 1, characterized in that, Obtaining a weld seam feature of the to-be-tested steel bar according to each pair of steel bar point set and remaining steel bar point set, specifically comprising: Based on a PCA algorithm, obtaining four points from each pair of steel bar point set and remaining steel bar point set; the four points are respectively two farthest points in the steel bar point set and two farthest points in the remaining steel bar point set; Based on a least square method, obtaining projection coordinates of the four points; Projecting the projection coordinates of the four points onto any central axis, and obtaining the welding feature of the welding seam area based on two points of the welding seam area.

5. The method for point cloud based reinforcement cross-weld and weld feature synchronous extraction according to claim 1, characterized in that, According to each pair of steel bar point sets and the remaining steel bar point sets, a final steel bar point set is obtained, specifically including: Based on the PCA algorithm, four points are obtained according to each pair of steel bar point sets and the remaining steel bar point sets; the four points are two farthest points in the steel bar point set and two farthest points in the remaining steel bar point set; and the final steel bar point set includes the four points.

6. The method for point cloud based reinforcement cross-weld and weld feature synchronous extraction according to claim 1, characterized in that, The welding point space coordinates are obtained according to each two final steel bar point sets, specifically including: Cylinder fitting is performed on each final steel bar point set to obtain an axis, a radius and an axis point; The welding point space coordinates are obtained according to the axis point on any final steel bar point set and the axis point on another final steel bar point set which is closest in distance.

Images