Automatic data segmentation for geometric feature extraction from unorganized 3-D coordinate points

Abstract

A systematic approach is proposed to automatically extract geometric surface features from a point cloud composed of a set of unorganized three-dimensional coordinate points by data segmentation. The point cloud is sampled from the boundary surface of a mechanical component with arbitrary shape. The proposed approach is composed of three steps. In the first step, a mesh surface domain is reconstructed to establish an explicit topological relation among the discrete points. The topological adjacency is further optimized to recover the second order object geometry. In the second step, curvature-based border detection is applied on the irregular mesh to extract both sharp borders with tangent discontinuity and smooth borders with curvature discontinuity. Finally, the mesh patches separated by the extracted borders are grouped together in the third step. For objects with complex shape, a multilevel segmentation scheme is proposed for better results. The capability of the proposed approach is demonstrated using various point clouds having distinct characteristics. Integrated with state of art scanning devices, the developed segmentation scheme can support reverse engineering of high precision mechanical components. It has potential applications in a whole spectrum of engineering problems with a major impact on rapid design and prototyping, shape analysis, and virtual reality.