Reconstruction of an Object Shape from Multiple Incomplete Range Data Sets Using Convex Hulls

This paper proposes a method of integrating multiple range data sets to be used as a base for accurate 3-D object modeling. This method consists of four parts: reducing each shape data set to its convex hull, matching facets of one convex hull to those of the other, estimating a transformation matrix by correspondences, and putting these data sets together. By using the convex hulls, integration is robust even if the data sets have deficiencies caused by the concave parts of an object. A facet on one convex hull that corresponds to a facet on another can be obtained concisely by utilizing the surface normals of the facets.