Object recognition using a distributed cooperative vision system

Abstract

Describes a distributed cooperative method whereby a complicated recognition process is divided into a function based independent recognition algorithm. Each algorithm can determine a certain unknown factor of the object under the condition that some remaining unknown factors are fixed. By introducing a cooperation mechanism, these independent algorithms can be used concurrently to determine all of the unknown factors. This cooperation mechanism is obtained by designing each algorithm as a recursive process which tries to reduce differences between predictive features obtained from a model and the features sensed. This approach results in a vision system that is easily adapted to many applications. A prototype vision system which determines the shape, location, and orientation of objects is developed. Using this system, it is experimentally verified that the proposed method is effective.<>