Fast point cloud boundary extraction technique combined with improved particle swarm algorithm

Abstract

The invention discloses a fast point cloud boundary extraction method combined with an improved particle swarm algorithm. The method comprises the following steps: obtaining point cloud data, establishing a topological relation and carrying out smoothing processing; selecting optimal feature points of all local structures through the improved particle swarm algorithm; through random sample consensus RANSAC, keeping N optimal feature points, and returning an optimal model; and according to uniformity of point distribution in point cloud data point k-neighborhood, finding all boundary feature points in the optimal model and finishing boundary extraction. Compared with a conventional point cloud boundary extraction method, the extraction method in the invention reduces point cloud boundary extraction calculation amount and is high in efficiency; since inertia weight is introduced, influence of the previous speed on the current speed can be controlled; by adjusting the magnitude of the inertia weight, the group can be prevented from falling into the local optimal point, and calculation precision is improved; and the method is close to a real object and is better in boundary effect.

Description

technical field [0001] The invention relates to a boundary extraction technology of a point cloud model, in particular to an improved particle swarm algorithm combined with a random sampling algorithm (RANSAC) in the boundary extraction process to quickly and accurately extract edge contours, and belongs to the technical field of point cloud processing. Background technique [0002] In real life, the spatial distribution of the surface of most objects is irregular, the topological relationship between each data point is not clear, and the obtained point cloud data is scattered and disorderly, so the point cloud must be processed first to obtain useful information. Since the point cloud boundary contains important information needed to express surface features and hole features, the fast and accurate boundary extraction directly affects the effect of point cloud subsequent processing. How to quickly, accurately and effectively extract the boundaries of scattered point clouds ...

AI Technical Summary

Problems solved by technology

Emelyanov et al. used the triangular mesh generated by the greedy strategy as a basis to judge whether the sampling point is a boundary point, but tetrahedrons may be generated during the triangulation process. also more

MilroyM and Yang used the technology of solving the extreme value of curvature to extract the boundary. This technology can effectively extract the boundary points of the point cloud model with a relatively small curvature change, that is, a relatively smooth and flat surface. However, for the point cloud model with a large curvature change, the effect is Poor; Orriols et al. applied the least square method to boundary feature extraction, but the accuracy of the algorithm to extract the boundary is not high; Bendels et al. used the minimum spanning tree to detect point cloud boundary feature points, but further improvement is needed in the complete extraction of boundary lines

Zhang Xianying et al. first established a triangular mesh, and identified the boundary of the point cloud by judging whether the adjacent points of the sampling points pass through the grid edge closed curve, but the complexity of establishing a triangular mesh is relatively high.

According to the geometric distribution characteristics of the point cloud, Liu Liqiang et al. calculated the distance between the center of gravity and the farthest point for boundary extraction, but the calculation accuracy obtained was not good.

Images