A RULE BASED 3D OBJECT RECOGNITION SYSTEM

Abstract

One goal of computer vision is to give (:ompUterO human 1 i ke vi sumi c:apabi Ii tics so that machines can sense the environment in their- field o+ view , Understand what is being sensed and take ai:)propriate action. Three dimensional object recognition greately facilitates the identification of the anticipated objects which can later be analysed for various Impc::rtant. app].Irations , suc::h as cietmct ion of crocks in the automobile during manufactur:irig , ideerttification and count of objects on a conveyor belt , detection of an incomming aircraft and robotic vision. The real world objects are normally built up from sub parts. Since the previously developed rule based 3d object recognition systems , concentrate on objects without sub parts , the recognition of objects containing sub parts is an important land mark in the field of 3d object recognition. It is shown that recognition of objects with sub parts can be performed with robustness and alacrity , even with incomplete image data that is due to image processing limitations or occlusion. The objects have been formed by combinations of various edges at junctions. Set of rules for describing known objects have been developed. This approach drastically reduces the search space and pattern matching time. Unknown objects have been identified by rejecting bad hypothesis , partial instantiation of remaining uninhibited hypothesis and accomplishing full instantiation of mode.t descriptor rules for objects wiLhout sub parts and with sub parts.