A Study of Feature Line Extraction and Closed Frame Structure of a Stone Tool from Measured Point Cloud

This paper introduces a method to extract feature lines and produce the closed frame structure of a stone tool from a measured point cloud. At first, the curvatures of the point cloud with polynomial fitting are derived partially. Then, from the mean curvature, inner and outer potential feature points sets are classified. After that, the most densely-located feature points are clustered with the Euclidean minimum spanning tree (EMST) algorithm. Then, the densely-located feature points are thinned with the Gaussian smoothing method. In this paper, the potential feature points are constructed by the EMST algorithm. Through the result of the EMST algorithm, noise nodes can be removed, and the longest feature line can be connected with the node valence of the EMST. Finally, the inner feature line is stitched by the outer feature line with a threshold. As an application, the closed frame structure of stone tools can be converted into 2D illustration in any viewpoint.