Machine Vision: Three Generations of Commercial Systems

Abstract

Since 1980. machine vision systems for industrial applications have enjoyed a rapidly expanding market. The first generation machines are 2-D binary vision systems, patterned after the SRI Vision Module. These systems will soon be joined by a second generation, based on edge description techniques. Both the first and second generation systems are pattern recognition machines. Research in machine vision is leading towards vision systems that will be able to dynamically model the 3-D surface in a scene. This research will lead to a third seneration of vision systems which will provide a dramatic increase in capabilities over thc first two generations. This article describes these three generations of vision systmens. The algoritms, data structures and hardware architecture are presented for binary vision systems and edge-based vision systems. A framework is presented for the reserach problems which must be solved before a commercial vision system can be produced bsaed on dynamic 3-D Scene analysis techniques.