Feature-based recognition of triangulated arbitrary surfaces

A procedure is presented for recognizing an object from its range data. A set of object surface data points are segmented into triangular patches using a small number of knot points. These knot points are placed by the local adjustment technique. For the resulting triangulated surface model, high-level edge lines are extracted using unit normal vectors of the triangular patches. This set of high-level primitives is used to construct a graph structure for each object. Experimental results are presented for a set of objects with salient features