Evaluation of existing and new feature recognition algorithms: Part 1: Theory and implementation

Abstract

This is the first of two papers evaluating the performance of general-purpose feature detection techniques for geometric models. In this paper, six different methods are described to identify sets of faces that bound depression and protrusion faces. Each algorithm has been implemented and tested on eight components from the National Design Repository. The algorithms studied include previously published general-purpose feature detection algorithms such as the single-face inner-loop and concavity techniques. Others are improvements to existing algorithms such as extensions of the two-dimensional convex hull method to handle curved faces as well as protrusions. Lastly, new algorithms based on the three-dimensional convex hull, minimum concave, visible and multiple-face inner-loop face sets are described. These algorithms provide a basis for the comparative analysis that is the subject of the second paper.