Experiments are reported that are directed toward the development of a vision system for bin-picking rigid 3-D objects. A structured light system is described for 3-D sensing and feature extraction that attempts to deliver a rich 2. 5-D representation of the scene. Two matching paradigms are studied for drawing correspondence between structures in the scene representation and structures in models. The matching algorithms presented depend on adequate feature detection and encounter problems of scale for multiple-object environments. It is concluded that an additional layer of object segmentation is required for a practical system.
|已發佈 - 1987
ASJC Scopus subject areas