The main goal of computer vision system-finding an increasing number of applications in diverse areas-is to recover useful information about an object or a scene.In many applications it is desirable to have a general vision system capable of analyzing images that it sees.However,this task of image understanding has been proved to be an exceedingly difficult one in a general domain.Therefore,the most vision systems exploit the domanin-specific knowledge.This paper addresses some of the major issues in utilizing different type of knowledge such as matching and control strategies