Abstract
Abstract
This paper describes a systematic procedure for developing composite feature detection systems from six methods for detecting three-dimensional depression features. The six methods, proposed by the authors in earlier papers, correspond to all the possible ways of grouping faces together from the simplest to the most complex grouping. All the possible ways of combining the six feature detection methods are considered and arranged in a tree structure. The possible composites are reduced to 20, using a tree pruning technique based on the criteria that the features detected should be the same (i.e. consistent), irrespective of the ordering of the faces in the B-rep model and that all faces of the component should be detected (i.e. complete coverage). A test bed for these 20 composites has been developed, implemented, and tested using carefully selected components from the public domain. The performance of these 20 composites is evaluated on the basis of suitability of the features as input to a machining application with minimal or no additional geometric reasoning, thus enabling the most promising composites to be identified.