Reinforcing steel bar intersection point identification and feature detection method based on point cloud

Through the point cloud-based method, the intersection points and weld characteristics of steel bars are extracted, and the problems of low identification efficiency and poor accuracy in the prior art are solved, and efficient and accurate detection of steel bars weld characteristics is achieved, which is suitable for a variety of scenarios.

CN118941517APending Publication Date: 2024-11-12SUZHOU QUANSHI INTELLIGENT OPTOELECTRONICS CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410964146.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing reinforcement welding feature extraction methods have strong targetedness and are unable to extract cross-point and weld features at the same time. They are low in efficiency and low in automation. The recognition accuracy of the method based on deep learning of images is not high, and the data processing is complex, making it difficult to adapt to complex On-site environment.

Method used

Using point cloud-based steel bar intersection recognition and feature detection methods, the three-dimensional steel bar point cloud under the world coordinate system is obtained, and the three-dimensional steel bar point cloud is converted to the user coordinate system, histogram analysis and cylinder fitting are performed, intersection points and gaps are calculated, and intersection points are realized. and efficient identification of weld features.

Benefits of technology

It realizes simultaneous identification of intersection points and weld features, improves identification efficiency and accuracy, is suitable for scenarios with different accuracy requirements, has strong anti-interference ability, and is suitable for complex construction site environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941517A_ABST
    Figure CN118941517A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of welding detection, and relates to a point cloud-based reinforcing steel bar intersection point identification and feature detection method, which comprises the following steps of: 1) acquiring a reinforcing steel bar three-dimensional point cloud and a tool center point cloud under a world coordinate system; 2) converting the steel bar three-dimensional point cloud in the world coordinate system into a steel bar three-dimensional point cloud in a first user coordinate system; 3) acquiring a steel bar three-dimensional point cloud histogram under the first user coordinate system, and acquiring the position of an intersection point based on the steel bar three-dimensional point cloud histogram; 4) obtaining point clouds in the interval range of each intersection point, fitting two cylinders according to the point clouds, and obtaining a cylinder coefficient; 5) voting to select a cylinder axis direction which is perpendicular to the X axis and has the most directions; 6) calculating an intersection point on the axes of the two cylinders in each intersection point interval; 7) calculating the midpoint of the intersection point and a steel bar gap; and 8) judging whether detection succeeds or not according to the steel bar gap, and converting the midpoint into a world coordinate system after detection succeeds. The method can identify the characteristics of the intersection points and the weld joints of the reinforcing steel bars efficiently and precisely in real time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of welding detection, and relates to a steel bar intersection recognition and feature detection method, and in particular to a steel bar intersection recognition and feature detection method based on point cloud. Background Art

[0002] With the development of social and economic levels, automation and robotics are also developing rapidly. The automation upgrade of building construction technology has become an important trend in the development of the construction industry, and welding robots are widely used in welding environments to replace manual work. The existence of steel bars in buildings varies depending on the work scene. During the steel bar welding process, the main recognition features of the robot include cross welds based on traditional geometry and deep learning based on images to identify welds and weld features.

[0003] Among them, the rebar tying robot is a new type of construction robot, which can replace manual tying of rebars in a standard environment and improve construction efficiency. Most of the current rebar tying devices use mechanical mechanisms or contact sensors to identify the rebar tying position, and then use the robot to tying the wire at the tying point, and then continue to identify and tie the next tying point after completion.

[0004] The existing steel bar welding feature extraction mainly obtains the three-dimensional point cloud data of the steel bar surface through 3D laser scanning, calculates its spatial geometric information, and processes it to obtain the desired intersection features. The deep learning-based method can also well identify intersection and weld features, but it requires a large number of data sets for training and cannot cope well with the changing on-site conditions.

[0005] The main disadvantages of the existing technology are: 1. It is highly targeted. At present, the traditional geometric cross weld feature extraction is only applicable to the extraction of cross weld features. It is highly targeted and cannot extract weld features at the same time; 2. Weld feature extraction is relatively difficult. The spatial cross steel bar point cloud with weld features cannot be segmented based on traditional Euclidean clustering, and weld features cannot be extracted; 3. Traditional detection methods mainly rely on manual work, with low efficiency and relatively low degree of automation; 4. The method based on image deep learning has the problem that its stability and recognition rate cannot reach 100%, and because image recognition lacks depth information and is easily affected by background during pixel-level segmentation, the accuracy of identified welding features is not high. At the same time, training requires a large number of data sets, which is not suitable for on-site project development; 5. The method based on point cloud deep learning requires a large number of annotated data sets for model training. Data processing and annotation are extremely complex. In addition, the construction site environment is complex and noisy, and it is difficult to obtain point cloud data sets through large-scale scanning.

[0006] Therefore, in view of the defects existing in the above-mentioned prior art, it is necessary to develop a new point cloud-based steel bar intersection recognition and feature detection method. Summary of the invention

[0007] In order to solve the above-mentioned defects and shortcomings in the prior art, the present invention proposes a point cloud-based steel bar intersection recognition and feature detection method, which can identify steel bar intersections and weld features in real time, efficiently and with high precision.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A method for identifying and detecting steel bar intersections based on point clouds, characterized in that it comprises the following steps:

[0010] 1) Obtain the three-dimensional point cloud of the steel bar and the center point cloud of the tool in the world coordinate system;

[0011] 2) establishing a first user coordinate system and converting the three-dimensional point cloud of the steel bar in the world coordinate system into the three-dimensional point cloud of the steel bar in the first user coordinate system;

[0012] 3) obtaining a three-dimensional point cloud histogram of the steel bars in the first user coordinate system based on the three-dimensional point cloud of the steel bars in the first user coordinate system, and obtaining a position where an intersection exists based on the three-dimensional point cloud histogram of the steel bars in the first user coordinate system;

[0013] 4) based on the position of each intersection, a point cloud within the range of each intersection is obtained, and two cylinders are fitted according to the point cloud within the range of each intersection to obtain the cylinder system coefficient;

[0014] 5) Vote to select the axis direction of the cylinder that is perpendicular to the X-axis of the first user coordinate system and has the most directions;

[0015] 6) Calculate the intersection points on the two cylinder axes within each intersection interval;

[0016] 7) Calculate the midpoint of the intersection and the gap between the steel bars;

[0017] 8) Determine whether the detection is successful based on the gap between the steel bars. If the detection is successful, convert the midpoint to the world coordinate system.

[0018] In addition, the present invention also provides a method for identifying and detecting steel bar intersections based on point clouds, which is characterized by comprising the following steps:

[0019] 1) Obtain the three-dimensional point cloud of the steel bar and the center point cloud of the tool in the world coordinate system;

[0020] 2) establishing a first user coordinate system and converting the three-dimensional point cloud of the steel bar in the world coordinate system into the three-dimensional point cloud of the steel bar in the first user coordinate system;

[0021] 3) obtaining a three-dimensional point cloud histogram of the steel bars in the first user coordinate system based on the three-dimensional point cloud of the steel bars in the first user coordinate system, and obtaining a position where an intersection exists based on the three-dimensional point cloud histogram of the steel bars in the first user coordinate system;

[0022] 4) based on the position of each intersection, a point cloud within the range of each intersection is obtained, and two cylinders are fitted according to the point cloud within the range of each intersection to obtain the cylinder system coefficient;

[0023] 5) Vote to select the axis direction of the cylinder that is perpendicular to the X-axis of the first user coordinate system and has the most directions;

[0024] 6) establishing a second user coordinate system, and converting the three-dimensional point cloud of the steel bars in the first user coordinate system to the three-dimensional point cloud of the steel bars in the second user coordinate system;

[0025] 7) obtaining a three-dimensional point cloud histogram of the steel bars in the second user coordinate system based on the three-dimensional point cloud of the steel bars in the second user coordinate system and obtaining a steel bar intersection area based on the histogram of the steel bar point cloud in the second user coordinate system;

[0026] 8) Fit two cylinders according to the point cloud in each steel bar intersection area to obtain the longitudinal steel bars and the transverse steel bars;

[0027] 9) Calculating the steel bar intersection points and steel bar gaps between the longitudinal steel bars and the transverse steel bars;

[0028] 10) determining whether the Z value of the longitudinal reinforcement at the intersection of the reinforcements in the second user coordinate system is greater than the Z value of the transverse reinforcement at the intersection of the reinforcements in the second user coordinate system, if so, generating a track along the transverse reinforcement on the upper surface of the longitudinal reinforcement, otherwise, generating a track along the longitudinal reinforcement on the upper surface of the transverse reinforcement;

[0029] 11) judging whether the detection is successful according to the steel bar gap, and converting the trajectory from the second user coordinate system to the world coordinate system after the detection is successful.

[0030] Preferably, the first user coordinate system is established in step 2) as follows: the forward direction of the tool center point cloud is the positive direction of the X-axis, the direction in which the centroid of the tool center point cloud points to the centroid of the steel bar three-dimensional point cloud is the positive direction of the Z-axis, and the positive direction of the Y-axis is obtained by the cross product. The entire first user coordinate system conforms to the right-hand rule, and the first user coordinate system is established with the first point cloud of the tool center point cloud as the origin.

[0031] Preferably, the step 3) is specifically as follows: along the positive direction of the X-axis of the first user coordinate system, with stride as the step length, the number of steel bar three-dimensional point clouds in each step length is counted as num_i (i=1,2,3,…,I), that is, a histogram of the steel bar three-dimensional point cloud in the first user coordinate system is obtained; the average value of the number of steel bar three-dimensional point clouds is obtained If num_i is greater than the average value numM, it is considered that there are crossed steel bars within this step, that is, there is an intersection.

[0032] Preferably, the step 4) is specifically as follows: according to the three-dimensional point cloud histogram of the steel bars in the first user coordinate system, the range [boundX_min, boundX_max] of the initial position of each intersection on the X-axis of the first user coordinate system is obtained, N strides are extended forward and backward along the X-axis of the first user coordinate system, and the range of the point cloud near the intersection on the X-axis of the first user coordinate system is obtained [boundX_min-N*stride, boundX_max+N*stride], the X-axis of the first user coordinate system is used as the filtering axis, and [boundX_min-N*stride, boundX_max+N*stride] is used as the reserved range, and each intersection interval Zone_j (j=1,…,m) is obtained by using straight-through filtering. ; Perform cylinder fitting on the three-dimensional point cloud of the steel bars in a certain intersection interval Zone_j, and obtain the axis direction [vx, vy, vz], radius radius and a point (x, y, z) on the axis of the cylinder. Each intersection interval is fitted twice, which are recorded as cylinder system coefficients coeff1 = [vx1, vy1, vz1, x1, y1, z1, radius1], coeff2 = [vx2, vy2, vz2, x2, y2, z2, radius2], and record the axis direction of the cylinder perpendicular to the X-axis of the first user coordinate system. If there are no two cylinders in a certain intersection interval, it is determined that the initial search of the intersection interval has failed, and the point cloud of the next intersection interval is directly fitted with a cylinder until all intersection intervals are processed.

[0033] Preferably, the second user coordinate system is established in step 6) as follows: obtain the direction with the most directions of the cylinder axis, calculate its direction average as the Y axis of the second user coordinate system, the X axis of the second user coordinate system is the same as the X axis of the first user coordinate system, and according to the right-hand rule, the Z axis of the second user coordinate system is cross-multiplied to establish the second user coordinate system.

[0034] Preferably, the step 7) is specifically as follows: along the positive direction of the X-axis of the second user coordinate system, with stride as the step length, counting the number of three-dimensional point clouds of steel bars in each step length, and obtaining a histogram of the three-dimensional point clouds of steel bars in the second user coordinate system; according to the histogram of the three-dimensional point clouds of steel bars in the second user coordinate system, obtaining the initial range [boundX_min, boundX_max] on the X-axis of the intersection in the second user coordinate system, and extending N strides forward and backward, then the range of the X-axis of the intersection area in the second user coordinate system is [boundX_min-N*stride, boundX_max+N*stride], taking the X-axis of the second user coordinate system as the retained dimension, and [boundX_min-N*stride, boundX_max+N*stride] as the retained range, and using pass-through filtering to obtain the point cloud of each steel bar intersection area.

[0035] Preferably, the step 8) is specifically as follows: fitting a cylinder to the point cloud in each steel bar intersection area, obtaining the axial direction, radius and a point on the axis of the cylinder, fitting twice in total, and recording the axial direction of the cylinder collinear with the X-axis of the second user coordinate system as the longitudinal steel bar, the radius of the longitudinal steel bar as radiusV, the axial direction of the longitudinal steel bar as [vxV, vyV, vzV], and a point PV(xV, yV, zV) on the axis of the longitudinal steel bar; recording the axial direction of the cylinder collinear with the Y-axis of the second user coordinate system as the transverse steel bar, the radius of the transverse steel bar as radiusH, the axial direction of the transverse steel bar as [vxH, vyH, vzH], and a point PV(xH, yH, zH) on the axis of the transverse steel bar, and only when there is one transverse steel bar and one longitudinal steel bar at the same time is the search considered successful, otherwise it is recorded as a search failure, and the next steel bar intersection area is searched until all steel bar intersection areas are searched.

[0036] Compared with the prior art, the point cloud-based steel bar intersection recognition and feature detection method of the present invention has one or more of the following beneficial technical effects:

[0037] 1. The present invention can simultaneously realize the recognition of intersections and the detection of weld features with high efficiency and high precision.

[0038] 2. The overall operation of the present invention is simple, and it can be applied to both scenes with low precision requirements and scenes with high precision requirements.

[0039] 3. The present invention has stronger anti-interference ability and is also convenient for complex processing of the trajectory, such as requiring the welding trajectory to be translated along a specified direction with higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention is a flow chart of the method for identifying and detecting steel bar intersections based on point clouds.

[0041] Figure 2 This is a statistical histogram of the steel bar three-dimensional point cloud of Example 1.

[0042] Figure 3 These are two steel bars fitted within a single intersection interval in Example 1.

[0043] Figure 4 This is the recognition result of Example 1, where white is the recognition point.

[0044] Figure 5 It is a histogram of the three-dimensional point cloud of the steel bars in the first user coordinate system of the second embodiment.

[0045] Figure 6 These are two steel bars fitted in the single steel bar intersection region of Example 2.

[0046] Figure 7 It is a histogram of the three-dimensional point cloud of the steel bars in the second user coordinate system of the second embodiment.

[0047] Figure 8 This is the welding trajectory of the identification point in the second embodiment. DETAILED DESCRIPTION

[0048] Before describing in detail any embodiment of the present invention, it should be understood that the present invention is not limited in its application to the construction and arrangement details of the components described below or illustrated in the following figures. The present invention can have other embodiments and can be practiced or carried out in various ways. In addition, it should be understood that the words and terms used here are for descriptive purposes and should not be considered restrictive. The use of "including" or "having" and its variations herein is intended to cover the items and their equivalents and additional items displayed below. Unless otherwise specified or limited, the terms "install", "connect", "support" and "couple" and their variations are widely used and cover direct installation and indirect installation, connection, support and connection. In addition, "connect" and "couple" are not limited to physical or mechanical connections or connections.

[0049] Furthermore, in the disclosure of the present invention, the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0050] Unless otherwise specified, the length unit used for the data in the present invention is millimeter (mm), the length unit in the image is pixel, the direction vector has no unit, diameter is the steel bar diameter, and hole is the sieve aperture.

[0051] Figure 1FIG. 4 is a flow chart showing the method for identifying and detecting steel bar intersections based on point clouds of the present invention. Figure 1 As shown, the steel bar intersection recognition and feature detection method based on point cloud of the present invention comprises the following steps:

[0052] 1. The robot drives the 3D laser sensor to scan the steel bars, obtain the steel bar 3D point cloud and tcp [tool center point] point cloud, and the coordinate system at this time is recorded as the world coordinate system.

[0053] The three-dimensional point cloud of the steel bar can be downsampled and outliers can be removed to obtain the steel bar point cloud after removing noise points.

[0054] 2. Take the forward direction of the tcp point cloud as the positive direction of the X-axis, the direction in which the centroid of the tcp point cloud points to the centroid of the steel bar point cloud as the positive direction of the Z-axis, and the positive direction of the Y-axis is obtained by the cross product. The entire coordinate system conforms to the right-hand rule. Take the first point of the tcp point cloud as the origin and establish the first user coordinate system T1.

[0055] Convert the 3D point cloud of the reinforcement in the world coordinate system to the 3D point cloud of the reinforcement in the first user coordinate system.

[0056] 3. Along the positive direction of the X-axis of the first user coordinate system, with stride as the step length, count the number of steel bar point clouds in each step length and record it as num_i (i=1,2,3,…,I), where I is the total number of steps, that is, obtain the steel bar point cloud histogram.

[0057] Get the average value of the number of steel bar point clouds If num_i is greater than the average value numM, it is considered that there may be cross-bars in the step interval, that is, there may be intersections. The step intervals that are continuously greater than the mean are merged and regarded as an intersection interval.

[0058] 4. According to the histogram, obtain the X-axis range [boundX_min, boundX_max] of the initial position of each intersection, extend N strides forward and backward along the X-axis of the first user coordinate system, and obtain the X-axis range [boundX_min-N*stride, boundX_max+N*stride] of the point cloud near the intersection. Take the X-axis as the filter axis and [boundX_min-N*stride, boundX_max+N*stride] as the retention range, and use straight-through filtering to obtain the point cloud within the range of each intersection interval Zone_j (j=1,…,m).

[0059] Fit a cylinder (i.e., steel bar) to the point cloud of a certain interval Zone_j, and obtain the axis direction [vx, vy, vz], radius radius, and a point (x, y, z) on the axis of the cylinder. Fit each interval twice, recorded as cylinder coefficients coeff1 = [vx1, vy1, vz1, x1, y1, z1, radius1], coeff2 = [vx2, vy2, vz2, x2, y2, z2, radius2], and record the axis direction of the cylinder perpendicular to the X-axis.

[0060] In the present invention, the specific fitting method is not required, such as the least squares fitting can be used. In addition, the two cylinder fitting processes are as follows: firstly, the first cylinder fitting is performed on all point clouds (cloudAll) within [boundX_min-N*stride, boundX_max+N*stride] to obtain the point cloud cloudIn1, the cylinder coefficient coeff1 and the remaining point cloud cloudRemain (the point cloud in cloudAll except cloudIn1) in the first cylinder; then the second cylinder fitting is performed on the point cloud of cloudRemain to obtain the cylinder coefficient coeff2.

[0061] If there are no two cylinders in a certain intersection interval, it is considered that the initial search for the intersection interval has failed, and the point cloud of the next intersection interval is directly processed (fitting the cylinder) until all intersection intervals are processed.

[0062] 5. Use the voting method to obtain the direction in which the axis of the cylinder in all intersection intervals is perpendicular to the X-axis and faces the most directions.

[0063] For rebar bundling, the intersection point is defined as the midpoint between the upper and lower bars and is characterized by the gap between the two crossing bars.

[0064] 6. Calculate the intersection of the two cylinder axes in each intersection interval. Given the two cylinder coefficients coeff1 and coeff2 in each intersection interval, the problem is now transformed into finding the intersection of two known spatial lines (cylinder axes) L1 and L2. Because the two spatial lines do not intersect, the projection of the virtual intersection on the two spatial lines is calculated as the intersection of the two spatial lines on each line [the specific method is not required, as long as it can be calculated].

[0065] 7. Calculate the midpoint of the intersection and the gap between the steel bars. The intersection point p3 (x3, y3, z3) on the straight line L1 and the intersection point p4 (x4, y4, z4) on the straight line L2 are identified by the midpoint of the two intersection points, that is, P = (p3 + p4) / 2 = ((x3 + x4) / 2, (y3 + y4) / 2, (z3 + z4) / 2).

[0066] At the same time, the distance between the two cylinders, i.e. the gap, is calculated. When the gap is larger than a certain value, it is considered that the workpiece assembly error is too large and subsequent processing is difficult.

[0067] 8. The following definition is made for whether the intersection is found successfully. If the intersection is found successfully, the identification point is output; if it is not found successfully, the identification point is not output:

[0068]

[0069] The identification point is the midpoint in step 7. The midpoint is converted to the world coordinate system. W =T1 -1 ·P, that's it.

[0070] For steel bar welding, the intersection is defined as a trajectory composed of multiple points, characterized by the gap between two crossing steel bars. In order to facilitate subsequent calculations and to ensure higher recognition accuracy, a second user coordinate system needs to be established.

[0071] 9. Get the direction with the most axes of the cylinder, calculate the average value of the direction as the Y axis of the second user coordinate system. The X axis of the second user coordinate system is the same as the X axis of the first coordinate system. According to the right-hand rule, cross multiply the Z axis to establish the second user coordinate system T2.

[0072] The steel bar 3D point cloud in the first user coordinate system is converted to the steel bar 3D point cloud in the second user coordinate system.

[0073] 10. Along the positive direction of the X-axis of the second user coordinate system, with stride as the step length, count the number of steel bar point clouds in each step length, and obtain the histogram of the number of steel bar point clouds in the second user coordinate system.

[0074] According to the histogram, get the initial X-axis range of the intersection point [boundX_min, boundX_max], expand N strides forward and backward, then the X-axis range of the intersection area is [boundX_min-N*stride, boundX_max+N*stride]. Take the X-axis as the reserved dimension and [boundX_min-N*stride, boundX_max+N*stride] as the reserved range, and use pass-through filtering to obtain the point cloud of each intersection interval.

[0075] 11. Fit the cylinder (that is, the steel bar) to the point cloud in each intersection interval, obtain the axis direction, radius and a point on the axis of the cylinder, and fit twice (fit the first one first, and then fit the second one in the remaining point cloud). The axis direction of the longitudinal steel bar is collinear with the X-axis, the radius of the longitudinal steel bar is radiusV, the axis direction of the longitudinal steel bar is [vxV, vyV, vzV], and the point on the axis of the longitudinal steel bar is PV (xV, yV, zV); the axis direction of the transverse steel bar is collinear with the Y-axis, the radius of the transverse steel bar is radiusH, the axis direction of the transverse steel bar is [vxH, vyH, vzH], and the point on the axis of the transverse steel bar is PV (xH, yH, zH). Only when there is one transverse steel bar and one longitudinal steel bar at the same time, the search is considered successful, otherwise it is recorded as a search failure, and the next intersection interval is searched until all intersection intervals are searched.

[0076] 12. If the search is successful, calculate the intersection points p5 (x5, y5, z5) and p6 (x6, y6, z6) of the two spatial axes (that is, the axes of the cylinder). The intersection points are on the axes of the horizontal and longitudinal reinforcements, respectively.

[0077] Simultaneously calculate the distance between two reinforcement bars (that is, two spatial axes), i.e. the gap When the gap is greater than a certain value, it is considered that the workpiece assembly error is too large and it is difficult to perform subsequent processing. Therefore, the following definition is made for whether the intersection is successfully found. When the intersection is successfully found, the identification point is output; if it is unsuccessful, the identification point is not output:

[0078]

[0079] 13. Given the intersection points p5 (x5, y5, z5) and p6 (x6, y6, z6) on the axis of the horizontal reinforcement and the axis of the longitudinal reinforcement, if z5 is greater than z6, an identification track is generated on the upper surface of the horizontal reinforcement along the direction of the longitudinal reinforcement; if z5 is less than z6, an identification track is generated on the upper surface of the longitudinal reinforcement along the direction of the horizontal reinforcement. If z5 is equal to z6, the search is considered to have failed.

[0080] If z5 is greater than z6, the direction vector vo is from the intersection p5 to p6, vo = normalize([x6-x5,y6-y5,z6-z5]) = [vox,voy,voz]; the identified trajectory points are composed of three points q1(xq1,yq1,zq1), q2(xq2,yq2,zq2), q3(xq3,yq3,zq3); Dis is defined as the spatial distance between q1 and q2, which can be customized by the user and the default is 5mm, then q1, q2, q3 can be calculated by the following formula.

[0081] xq1=x5+vox*radiusH+vxH*radiusV-vxV*Dis;

[0082] yq1=y5+voy*radiusH+vyH*radiusV-vyV*Dis;

[0083] zq1=z5+voz*radiusH+vzH*radiusV-vzV*Dis;

[0084] xq2=x5+vox*radiusH+vxH*radiusV;

[0085] yq2=y5+voy*radiusH+vyH*radiusV;

[0086] zq2=z5+voz*radiusH+vzH*radiusV;

[0087] xq3=x5+vox*radiusH+vxH*radiusV+vxV*Dis;

[0088] yq3=y5+voy*radiusH+vyH*radiusV+vyV*Dis;

[0089] zq3=z5+voz*radiusH+vzH*radiusV+vzV*Dis;

[0090] If z5 is less than z6, the direction vector vo points from the intersection p6 to p5, vo = normalize([x5-x6,y5-y6,z5-z6]) = [vox,voy,voz]; then the three generated points q1(xq1,yq1,zq1), q2(xq2,yq2,zq2), q3(xq3,yq3,zq3) can be calculated by the following formula.

[0091] xq1=x6+vox*radiusV+vxV*radiusH-vxH*Dis;

[0092] yq1=y6+voy*radiusV+vyV*radiusH-vyH*Dis;

[0093] zq1=z6+voz*radiusV+vzV*radiusH-vzH*Dis;

[0094] xq2=x6+vox*radiusV+vxV*radiusH;

[0095] yq2=y6+voy*radiusV+vyV*radiusH;

[0096] zq2=z6+voz*radiusV+vzV*radiusH;

[0097] xq3=x6+vox*radiusV+vxV*radiusH+vxH*Dis;

[0098] yq3=y6+voy*radiusV+vyV*radiusH+vyH*Dis;

[0099] zq3=z6+voz*radiusV+vzV*radiusH+vzH*Dis;

[0100] 14. Convert the calculated recognition trajectory from the second user coordinate system to the world coordinate system Q W =T1 -1 T2 -1 ·q.

[0101] In the present invention, the step length stride in the histogram statistics is 10 mm by default, and the recommended range is [8,12] mm; the number of forward and backward extension steps N is 5 by default, and the recommended range is [3,6].

[0102] The following describes in detail the point cloud-based steel bar intersection recognition and feature detection method of the present invention with several specific embodiments, so that those skilled in the art can better understand the present invention.

[0103] Example 1: Steel bar binding

[0104] 1. Obtain the steel bar 3D point cloud and TCP point cloud in the world coordinate system.

[0105] 2. Downsample and remove outliers on the three-dimensional point cloud of steel bars to obtain the steel bar point cloud after removing noise points.

[0106] 3. Get the forward direction vx of the tcp point cloud = [-1.00, 0.00, 0.00], the center of mass of the steel bar point cloud cp1 (-238.38, -2167.96, 847.51), and establish the first user coordinate system

[0107] 4. Convert the steel bar point cloud in the world coordinate system to the first user coordinate system to obtain point cloud Cloud1.

[0108] 5. Along the positive direction of the X-axis of the first user coordinate system, with stride = 10 mm as the step length, count the number of steel bar point clouds in each step length, and obtain the histogram of the number of steel bar point clouds as shown below: Figure 2shown.

[0109] 6. From the histogram, we can see that the intervals greater than the mean value are the locations where the intersections may exist. Take N equal to 5, and obtain the X-axis range of each interval in the first user coordinate system, see Table 1.

[0110] 7. Using straight-through filtering, take the point cloud of each interval along the X-axis, and perform two cylinder fittings in the interval. The two steel bars of a single interval fitting are shown in Figure 1. Figure 3 Obtain the axis direction, radius and a point on the axis of the cylinder twice, and record the axis direction of the cylinder perpendicular to the X-axis, see Table 1.

[0111] 8. Use the voting method to select the direction with the most orientations, and eliminate the other directions. The selected ones are used to calculate the intersection of two spatial lines in each interval, and the midpoint of the intersection is calculated as the identification point, which is then converted to the world coordinate system, and the gaps between the steel bars are calculated at the same time. The results are shown in Table 1. The gaps all meet the output requirements, so they are all output as valid identification points. Among them, the recognition results of this embodiment are shown in Table 1. Figure 4 , white is the identification point.

[0112] Table 1: Key calculation results of Example 1

[0113]

[0114]

[0115] Example 2: Steel bar welding, main steel bar on top

[0116] 1. Obtain the steel bar 3D point cloud and TCP point cloud in the world coordinate system.

[0117] 2. Downsample and remove outliers on the three-dimensional point cloud of steel bars to obtain the steel bar point cloud after removing noise points.

[0118] 3. Get the forward direction vx of the tcp point cloud = [-0.99, 0.10, 0.09], the center of mass of the steel bar point cloud cp1 (-246.703, -2149.490, 838.971), and establish the first user coordinate system

[0119] 4. Convert the steel point cloud in the world coordinate system to the first user coordinate system to obtain point cloud Cloud1

[0120] 5. Along the positive direction of the X-axis of the first user coordinate system, with stride = 10 mm as the step length, count the number of steel bar point clouds in each step length, and obtain the histogram of the number of steel bar point clouds as shown below: Figure 5 .

[0121] 6. From the histogram, we can see that the intervals where the number of point clouds is greater than the mean are possible intersections. Take N equal to 5, and get the X coordinate range of each interval as shown in Table 2.

[0122] 7. Using straight-through filtering, take the point cloud of each interval along the X-axis, and perform two cylinder fittings in the interval. The two steel bars of a single interval fitting are as follows: Figure 6 Obtain the axis direction, radius and a point on the axis of the cylinder twice, and record the axis direction of the cylinder perpendicular to the X-axis, see Table 2.

[0123] Table 2: First user coordinate system interval information of Example 2

[0124] Interval Interval range X coordinate (mm) Axis direction perpendicular to the X-axis Is it the most oriented Interval 1 [-19.083,100.917] [-0.011,0.751,-0.660] yes Interval 2 [130.917,250.917] [0.000,0.738,-0.675] yes Interval 3 [290.917,410.917] [-0.005,0.753,-0.658] yes

[0125] 6. Use the voting method to select the direction with the most orientations, and take the average of the directions with the most orientations as the final selected steel bar axis direction, recorded as vy = [0.00, 0.747, -0.665]. Take the positive direction of the X axis of the first user coordinate system as vx, cross multiply to get the Z axis direction, and establish the second user coordinate system

[0126]

[0127] 7. Convert the steel bar point cloud cloud1 in the first user coordinate system to the second user coordinate system to obtain point cloud cloud2.

[0128] 8. Using straight-through filtering, take the point cloud of each interval along the X-axis and get the histogram as follows Figure 7 As shown. Two cylinder fittings are performed in this interval to obtain the parameters and gaps of the horizontal and longitudinal reinforcements. From the gap, it can be seen that the gap in the second interval is greater than the set threshold of 8mm, so this interval is considered to have failed.

[0129] 9. Find the final generated trajectory for intervals 1 and 3, such as Figure 8For interval 1, because the Z coordinate value of the intersection point on the longitudinal steel bar axis is smaller than the Z coordinate value of the intersection point on the transverse steel bar axis, the generated welding trajectory is on the upper surface of the transverse steel bar, pointing to the vector vo = normalize ([42.286, -133.019, 150.987] - [42.272, -133.104, 161.821]) = normalize ([0.014, 0.085, -10.834]) = [0.001, 0.008, -0.999]. Taking Dis as 5mm, the three points generated in the second user coordinate system are [37.3254,-138.261,155.493], [42.3252,-138.225,155.5], [47.3251,-138.188,155.507], and converted to the world coordinate system are [-101.815,-2155.46,844.289], [-106.814,-2155.47,844.256], [-111.814,-2155.47,844.223]. Similarly, the coordinates of the three points in interval 3 in the world coordinate system are [-412.736,-2156.71,843.205], [-417.735,-2156.74,843.222], [-422.735,-2156.77,843.239].

[0130] Table 2: Example 2 in the second user coordinate system

[0131]

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Those skilled in the art can modify or replace the technical solution of the present invention according to the idea of ​​the present invention without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A method for identifying and detecting steel bar intersections based on point cloud, characterized in that: The following steps are involved: 1) Obtain the three-dimensional point cloud of the steel bar and the center point cloud of the tool in the world coordinate system; 2) establishing a first user coordinate system and converting the three-dimensional point cloud of the steel bar in the world coordinate system into the three-dimensional point cloud of the steel bar in the first user coordinate system; 3) obtaining a three-dimensional point cloud histogram of the steel bars in the first user coordinate system based on the three-dimensional point cloud of the steel bars in the first user coordinate system, and obtaining a position where an intersection exists based on the three-dimensional point cloud histogram of the steel bars in the first user coordinate system; 4) based on the position of each intersection, a point cloud within the range of each intersection is obtained, and two cylinders are fitted according to the point cloud within the range of each intersection to obtain the cylinder system coefficient; 5) Vote to select the cylinder axis direction that is perpendicular to the X-axis of the first user coordinate system and has the most directions; 6) Calculate the intersection points on the two cylinder axes within each intersection interval; 7) Calculate the midpoint of the intersection and the gap between the steel bars; 8) Determine whether the detection is successful based on the gap between the steel bars. If the detection is successful, convert the midpoint to the world coordinate system.

2. A method for identifying and detecting steel bar intersections based on point cloud, characterized in that: The following steps are involved: 1) Obtain the three-dimensional point cloud of the steel bar and the center point cloud of the tool in the world coordinate system; 2) establishing a first user coordinate system and converting the three-dimensional point cloud of the steel bar in the world coordinate system into the three-dimensional point cloud of the steel bar in the first user coordinate system; 3) obtaining a three-dimensional point cloud histogram of the steel bars in the first user coordinate system based on the three-dimensional point cloud of the steel bars in the first user coordinate system, and obtaining a position where an intersection exists based on the three-dimensional point cloud histogram of the steel bars in the first user coordinate system; 4) based on the position of each intersection, a point cloud within the range of each intersection is obtained, and two cylinders are fitted according to the point cloud within the range of each intersection to obtain the cylinder system coefficient; 5) Vote to select the cylinder axis direction that is perpendicular to the X-axis of the first user coordinate system and has the most directions; 6) establishing a second user coordinate system, and converting the three-dimensional point cloud of the steel bars in the first user coordinate system to the three-dimensional point cloud of the steel bars in the second user coordinate system; 7) obtaining a three-dimensional point cloud histogram of the steel bars in the second user coordinate system based on the three-dimensional point cloud of the steel bars in the second user coordinate system and obtaining a steel bar intersection area based on the histogram of the steel bar point cloud in the second user coordinate system; 8) Fit two cylinders according to the point cloud in each steel bar intersection area to obtain the longitudinal steel bars and the transverse steel bars; 9) Calculating the steel bar intersection points and steel bar gaps between the longitudinal steel bars and the transverse steel bars; 10) determining whether the Z value of the longitudinal reinforcement at the intersection of the reinforcements in the second user coordinate system is greater than the Z value of the transverse reinforcement at the intersection of the reinforcements in the second user coordinate system, if so, generating a track along the transverse reinforcement on the upper surface of the longitudinal reinforcement, otherwise, generating a track along the longitudinal reinforcement on the upper surface of the transverse reinforcement; 11) judging whether the detection is successful according to the steel bar gap, and converting the trajectory from the second user coordinate system to the world coordinate system after the detection is successful.

3. The method for identifying and detecting steel bar intersections based on point cloud according to claim 1 or 2, characterized in that: The first user coordinate system is established in step 2) as follows: the forward direction of the tool center point cloud is the positive direction of the X-axis, the direction from the center of mass of the tool center point cloud to the center of mass of the steel bar three-dimensional point cloud is the positive direction of the Z-axis, and the positive direction of the Y-axis is obtained by the cross product. The entire first user coordinate system conforms to the right-hand rule, and the first point cloud of the tool center point cloud is used as the origin to establish the first user coordinate system.

4. The method for identifying and detecting steel bar intersections based on point cloud according to claim 3, characterized in that: The step 3) is specifically as follows: along the positive direction of the X-axis of the first user coordinate system, with stride as the step length, the number of three-dimensional point clouds of steel bars in each step length is counted and recorded as num_i (i=1,2,3,…,I), that is, a histogram of the three-dimensional point clouds of steel bars in the first user coordinate system is obtained; and the average number of the three-dimensional point clouds of steel bars is obtained. If num_i is greater than the average value numM, it is considered that there are crossed steel bars within this step, that is, there is an intersection.

5. The method for identifying and detecting steel bar intersections based on point cloud according to claim 4, characterized in that: The step 4) is specifically as follows: according to the three-dimensional point cloud histogram of the steel bars in the first user coordinate system, the range [boundX_min, boundX_max] of the initial position of each intersection on the X-axis of the first user coordinate system is obtained, N strides are extended forward and backward along the X-axis of the first user coordinate system to obtain the range [boundX_min-N*stride, boundX_max+N*stride] of the point cloud near the intersection on the X-axis of the first user coordinate system, the X-axis of the first user coordinate system is used as the filtering axis, and [boundX_min-N*stride, boundX_max+N*stride] is used as the reserved range, and the point cloud in each intersection interval Zone_j (j=1,…,m) is obtained by using straight-through filtering. Rebar 3D point cloud; perform cylinder fitting on the 3D point cloud of the rebar in a certain intersection interval Zone_j, obtain the axis direction [vx, vy, vz], radius radius and a point on the axis (x, y, z) of the cylinder, and fit each intersection interval twice, which are recorded as cylinder system coefficients coeff1 = [vx1, vy1, vz1, x1, y1, z1, radius1], coeff2 = [vx2, vy2, vz2, x2, y2, z2, radius2], and record the axis direction of the cylinder perpendicular to the X-axis of the first user coordinate system. If there are no two cylinders in a certain intersection interval, it is determined that the initial search of the intersection interval has failed, and the point cloud of the next intersection interval is directly fitted with a cylinder until all intersection intervals are processed.

6. The method for identifying and detecting steel bar intersections based on point cloud according to claim 2, characterized in that: The second user coordinate system is established in step 6) as follows: obtain the direction with the most directions of the cylinder axis, calculate its average direction as the Y axis of the second user coordinate system, the X axis of the second user coordinate system is the same as the X axis of the first user coordinate system, and according to the right-hand rule, the Z axis of the second user coordinate system is cross-multiplied to establish the second user coordinate system.

7. The method for identifying and detecting steel bar intersections based on point cloud according to claim 6, characterized in that: The step 7) is specifically as follows: along the positive direction of the X-axis of the second user coordinate system, with stride as the step length, the number of three-dimensional point clouds of steel bars in each step length is counted, and a histogram of the three-dimensional point clouds of steel bars in the second user coordinate system is obtained; according to the histogram of the three-dimensional point clouds of steel bars in the second user coordinate system, the initial range [boundX_min, boundX_max] on the X-axis of the intersection in the second user coordinate system is obtained, and N strides are extended forward and backward, then the range of the X-axis in the second user coordinate system of the intersection area is [boundX_min-N*stride, boundX_max+N*stride], the X-axis of the second user coordinate system is used as the reserved dimension, and [boundX_min-N*stride, boundX_max+N*stride] is used as the reserved range, and the point cloud of each steel bar intersection area is obtained by using straight-through filtering.

8. The method for identifying and detecting steel bar intersections based on point cloud according to claim 7, characterized in that: The step 8) is specifically as follows: fitting a cylinder for the point cloud in each steel bar intersection area, obtaining the axis direction, radius and a point on the axis of the cylinder, fitting twice in total, recording the axis direction of the cylinder collinear with the X-axis of the second user coordinate system as the longitudinal steel bar, the radius of the longitudinal steel bar as radiusV, the axis direction of the longitudinal steel bar as [vxV, vyV, vzV], and a point PV(xV, yV, zV) on the axis of the longitudinal steel bar; recording the axis direction of the cylinder collinear with the Y-axis of the second user coordinate system as the transverse steel bar, the radius of the transverse steel bar as radiusH, the axis direction of the transverse steel bar as [vxH, vyH, vzH], and a point PV(xH, yH, zH) on the axis of the transverse steel bar, and only when there is one transverse steel bar and one longitudinal steel bar at the same time is the search considered successful, otherwise it is recorded as a search failure, and the next steel bar intersection area is searched until all steel bar intersection areas are searched.

Citation Information

Cited By

  • Linear welding seam identification method and device

    CN119131037A

  • Welding spot positioning method and device, computer equipment, readable storage medium and program product

    CN120876585A

  • Nuclear fuel rod gap measurement method, device and equipment and storage medium

    CN122175850A