A Direct Segmentation Algorithm Based on Adaptive Fuzzy c-Elliptotype Clustering in Reverse Engineering

Reverse engineering is the process of converting a large measured point cloud into a CAD model. The key issue is segmentation, i.e. how to subdivide the point cloud into smaller regions, where each of which can be approximated by a single surface. This paper presents segmentation algorithm based on adaptive fuzzy c-elliptotype clustering for point cloud in reverse engineering. Fuzzy c-planes clustering is added to the classic fuzzy c-means algorithm by convex combination. The combination coefficient is selected adaptively according to size of planes by a heuristic technique. The differential geometric features that are necessary in the conventional segmentation algorithm do not need to be computed. The number of clusters is determined by competitive agglomeration technique. The algorithm is benefit to extract geometric parameters of surface accurately in next step. The experiment results prove that this algorithm is valid.