Abstract: As the most important geometry feature of CAD model in reverse engineering, how to accurately extract the boundary of point clouds is the key to guarantee the quality and precision of the final reconstructed surface. The mathematical model of boundary extraction is constructed firstly through the segmentation of the spatial box bounded by arbitrary uneven point clouds. Secondly, the seed recognition process and the seed growing process are implemented respectively to separate the boundary regions from the scattered point set by simple topology operation of 3-D grids. The feature points of boundary are extracted finally by applying topological-graph reasoning based on optimization of intrinsic geometry attribute. Empirical results show that the proposed algorithm is afficient and reliable in practice.

Key words: Boundary extraction, Point clouds, Spatial grid, Topological graph