A Dempster-Shafer approach for recognizing machine features from CAD models

This paper introduces an evidential reasoning-based approach for recognizing and extracting manufacturing features from solid model description of objects. A major difficulty faced by previously proposed methods for feature extraction has been the interaction between features due to non-uniqueness and ambiguousness in feature representation. To overcome this difficulty, we introduce a Dempster-Shafer approach for generating and combining geometric and topologic evidences to identify and extract interacting features. The main contributions of this research include introducing different classes of evidences based on the geometric and topologic relationships at different abstraction levels for effective evidential reasoning and developing the principle of association to overcome the mutual exclusiveness assumption of the Dempster-Shafer theory. Experiments demonstrate the effectiveness of the proposed approach in extracting interacting machine features.