CN116862935A - Automatic extraction method for point cloud edge contour for cylindrical three-dimensional measurement - Google Patents

Abstract

The invention provides a method for automatically extracting point cloud edge contours for cylindrical three-dimensional measurement, which includes the following steps: S1. Use a three-dimensional laser scanning device to obtain cylindrical three-dimensional scanning point cloud data and input it; S2. First-order point cloud data Segmentation; S3, establish a graph connecting adjacent clusters; S4, add edges representing feature lines; S5, construct the minimum spanning tree of the pruned graph; S6, feature lines of the minimum spanning tree of the closed graph; S7, output contour feature lines: Use the plane curve evolution method to smooth the contour feature lines and output them. The present invention is different from the existing characteristic line algorithm by reconstructing the closed cylindrical characteristic line. The advantages of this algorithm are: it only uses the coordinates of the point cloud for adaptive calculation, without the need to frame the point cloud area, it is a gridless automatic feature extraction, and it can intelligently cluster the point cloud to form a smaller point cloud than the original point cloud. A lot of graphs improve the real-time performance of calculations.

Description

An automatic extraction method of point cloud edge contours for cylindrical three-dimensional measurement

Technical field

The invention relates to the field of machinery and automation technology, and in particular to an automatic point cloud edge contour extraction method for cylindrical three-dimensional measurement.

Background technique

At present, 3D laser scanning measurement technology has been increasingly used in the measurement of large-size parts such as cylindrical parts due to its characteristics of large measurement range, high precision, non-contact, not affected by light, fast scanning speed, and high degree of automation. Measurement. The dimensions that need to be measured for the cylindrical shape mainly include parameters such as the diameter, roundness, height, verticality, and reference deflection of the docking hole fitting circle. The measurement of these parameters requires picking up the edge contour of the cylindrical point cloud data and then using the edge contour. Point calculation to obtain the measurement results.

Existing object edge contour recognition methods mainly include detection based on two-dimensional images and feature point recognition based on three-dimensional point cloud data. Image detection methods have certain limitations because they cannot reflect the complete information of three-dimensional parts. The existing recognition methods based on three-dimensional point cloud data have the characteristics of large amount of calculation, poor adaptability of parameter values, and low measurement accuracy. In order to improve the recognition efficiency and accuracy of cylindrical edge contour point cloud data, the present invention proposes an automatic extraction method of cylindrical edge contour based on point cloud data.

Contents of the invention

The purpose of the present invention is to provide a point cloud edge contour automatic extraction method for cylindrical three-dimensional measurement, which realizes automatic extraction of edge contours by searching for closed feature points of cylindrical point cloud data, has fast running speed and good extraction effect.

The invention provides an automatic point cloud edge contour extraction method for cylindrical three-dimensional measurement, which includes the following steps:

S1. Use 3D laser scanning equipment to obtain cylindrical 3D scanning point cloud data and input it;

S2, first-order segmentation of point cloud data;

S3. Create a graph connecting adjacent clusters;

S4. Add edges representing feature lines;

S5. Construct the minimum spanning tree of the pruning graph;

S6. Characteristic lines of the minimum spanning tree of closed graphs;

S7. Output the contour feature line: Use the plane curve evolution method to smooth the contour feature line and output it.

Further, the step S2 is specifically:

By generating a sphere with p as the midpoint and radius, the k nearest neighbors are within the sphere, and the least square plane of these points is constructed for projection; the normal distribution estimation regional growth method is used to cluster the local area point cloud. Class, by setting the threshold angle to specify the maximum angle between the normals of two adjacent points in a cluster and complete the first-order normal vector segmentation of the point cluster.

Further, the step S3 is specifically:

A connected graph is constructed as an initial approximation of feature lines using a minimum spanning tree algorithm, where each vertex represents a cluster and each edge connects two clusters containing at least one point with overlapping neighborhoods, and the detection of prominent feature lines is completed.

Further, the step S4 is specifically:

In the connected graph constructed in step S3, add a characteristic line edge whose Euclidean distance is less than the Euclidean distance between two adjacent small clusters in the small cluster connecting two adjacent large clusters in the graph, and eliminate the edges caused by the prominent edges. The gap between clusters caused by the alignment of points with the normals of adjacent large clusters realizes the closure of the boundaries.

Further, the step S5 is specifically:

Calculate the weight of the edges between small clusters as the distance between the representative points of the clusters, and attach the weights greater than those between small clusters to the connected edges of large clusters to obtain a graph with a reduced number of edges. At this time, there are only limited The edge involves a large cluster, completing the construction of a minimum spanning tree and the pruning of short branches.

Further, the step S6 is specifically:

Using a connection algorithm with a depth of path optimization search between two endpoints, each endpoint in the graph is connected to an appropriate point to form a closed edge feature line.

The beneficial effects achieved by the point cloud edge contour automatic extraction method for cylindrical three-dimensional measurement provided by the present invention are:

1) The present invention is different from the existing characteristic line algorithm by reconstructing the closed cylindrical characteristic line. The advantages of this algorithm are: it only uses the coordinates of the point cloud for adaptive calculation, without the need to frame the point cloud area, it is a gridless automatic feature extraction, and it can intelligently cluster the point cloud to form a smaller point cloud than the original point cloud. A lot of graphs improve the real-time performance of calculations.

2) The graph structure based on point clusters constructed by the present invention extracts feature lines on the coordinates of a single point, which can well identify the prominent edges of high normal vectors, easily segment the positions of these prominent edges, and reduces the calculation time. complexity.

3) On the basis of obtaining the contour points, the present invention uses the original position information of the local area points to connect adjacent points to form the initial contour line represented by the graph, and uses several steps such as minimum spanning tree, short edge pruning and smoothing. The processing forms the final smooth contour line, which has low computational complexity and high speed. It can realize automatic extraction of point cloud edge contours and can be easily embedded in measurement software to achieve one-click size measurement.

Description of the drawings

The invention will be further described below with reference to the accompanying drawings:

Figure 1 is a flow chart of the automatic extraction method of point cloud edge contours for cylindrical three-dimensional measurement;

Figure 2 is a flow chart of k neighborhood point selection;

Figure 3 is a flow chart of point cluster segmentation and connection closure;

Figure 4 is a minimum spanning tree flow chart for constructing a pruning graph.

Detailed ways

The automatic point cloud edge contour extraction method proposed by the present invention for cylindrical three-dimensional measurement will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the invention will become more apparent from the following description and claims. It should be noted that the drawings are in a very simplified form and use imprecise ratios, and are only used to conveniently and clearly assist in explaining the embodiments of the present invention.

Example 1

The present invention proposes an automatic edge contour extraction method based on three-dimensional point cloud data for cylindrical measurement. The edge contour is automatically extracted by searching for closed feature points of cylindrical point cloud data. The operation speed is fast and the extraction effect is good.

The main technical solution of the present invention is to construct a cylindrical curve network on the cylindrical point cloud, use the normal estimation region growth method to cluster the point cloud to form point clusters, further reduce the size of the point cloud, and establish a graph of these point clusters. And use first-order vector feature segmentation to extract closed feature lines that match the clusters, and find the area where the cylindrical surface contour can be defined.

The present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings.

Figure 1 is a flow chart of the point cloud edge contour automatic extraction method used for cylindrical three-dimensional measurement; Figure 2 is a k-neighbor point selection flow chart; Figure 3 is a point cluster segmentation and connection closed flow chart; Figure 4 is a pruning diagram construction Minimum spanning tree flow chart. Referring to Figures 1 to 4, the main steps of this embodiment are as follows:

Step S1: Use 3D laser scanning equipment to obtain cylindrical 3D scanning point cloud data and input it;

Step S2: First-order segmentation of point cloud data. By generating a sphere with p as the midpoint and radius, the k nearest neighbors are within the sphere, and the least square plane of these points is constructed for projection. Use the normal distribution estimation regional growth method to cluster local area point clouds, and set the threshold angle to specify the maximum angle between the normals of two adjacent points in a cluster and complete the first-order normal vector segmentation of the point cluster;

Step S3: Create a graph connecting adjacent clusters. Use the minimum spanning tree algorithm to construct a connected graph as an initial approximation of feature lines, where each vertex represents a cluster and each edge connects two clusters that contain at least one point and have overlapping neighborhoods, and complete the detection of prominent feature lines;

Step S4: Add edges representing feature lines. In the small cluster connecting two adjacent large clusters in the figure, add a characteristic line edge whose Euclidean distance is smaller than the Euclidean distance between the two adjacent small clusters in the figure, and eliminate the differences between the points on the prominent edges and the adjacent large clusters. The gaps between clusters caused by normal alignment achieve boundary closure;

Step S5: Construct the minimum spanning tree of the pruning graph. Calculate the weight of the edges between small clusters as the distance between the representative points of the clusters, and attach the weights greater than those between small clusters to the connected edges of large clusters to obtain a graph with a reduced number of edges. At this time, there are only limited The edge involves a large cluster, completing the construction of a minimum spanning tree and the pruning of short branches;

Step S6: Characteristic lines of the minimum spanning tree of the closed graph. Use a connection algorithm with a depth of path optimization search between two endpoints to connect each endpoint in the graph with an appropriate point to form a closed edge feature line;

Step S7: Output the contour feature lines. Use the plane curve evolution method to smooth the contour feature lines and output them.

Contents not described in detail in this specification belong to the prior art known to those skilled in the art. It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as illustrative and non-restrictive from any point of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is therefore intended that all claims falling within the claims All changes within the meaning and scope of equivalent elements are included in the present invention.

Claims (6)

1. The automatic extraction method for the edge contour of the point cloud for cylindrical three-dimensional measurement is characterized by comprising the following steps of:

s1, acquiring cylindrical three-dimensional scanning point cloud data by using a three-dimensional laser scanning device and inputting the data;

s2, first-order segmentation of point cloud data;

s3, establishing a graph for connecting adjacent clusters;

s4, adding edges representing the characteristic lines;

s5, constructing a minimum spanning tree of the pruning graph;

s6, closing a characteristic line of the minimum spanning tree of the graph;

s7, outputting contour feature lines: and smoothing the contour characteristic line by using a plane curve evolution method and outputting the contour characteristic line.

2. The method for automatically extracting the edge contour of the point cloud for cylindrical three-dimensional measurement according to claim 1, wherein the step S2 is specifically:

generating a spherical surface taking p as a midpoint and a radius, enabling k nearest neighbors to be in the spherical surface, and constructing a least square plane of the points for projection; clustering the local area point clouds by using a normal distribution estimation area growth method, designating the maximum angle between two adjacent point normal lines in one cluster by setting a threshold angle, and completing the first-order normal vector segmentation of the point clusters.

3. The method for automatically extracting the edge profile of the point cloud for cylindrical three-dimensional measurement according to claim 2, wherein the step S3 is specifically:

constructing a connected graph using a minimum spanning tree algorithm as an initial approximation of the feature line, wherein each vertex represents a cluster, each edge connects two clusters containing at least one point and having overlapping neighbors, and the detection of the salient feature line is completed.

4. The method for automatically extracting a point cloud edge profile for cylindrical three-dimensional measurement according to claim 3, wherein the step S4 specifically comprises:

and adding a characteristic line edge with the Euclidean distance smaller than that between two adjacent small clusters in the graph into the small clusters which are connected with the two adjacent large clusters in the connected graph constructed in the step S3, eliminating gaps among the clusters caused by alignment of points on the protruding edge and normal lines of the adjacent large clusters, and realizing boundary closure.

5. The method for automatically extracting the edge contour of the point cloud for cylindrical three-dimensional measurement according to claim 4, wherein the step S5 specifically comprises:

and calculating the weight of edges between small clusters, taking the weight as the distance between representative points of the clusters, and adding the weight larger than the weight between the small clusters to the edges of the connected large clusters to obtain a graph with reduced edge numbers, wherein only limited edges relate to the large clusters at the moment, and constructing the minimum spanning tree and pruning short branches.

6. The method for automatically extracting the edge contour of the point cloud for cylindrical three-dimensional measurement according to claim 5, wherein the step S6 is specifically:

each endpoint in the graph is connected with a suitable point to form a closed edge feature line by using a connection algorithm of path optimization search between two endpoints.