A Recognition Method Combining CAD Inforamtion with Image Processing Using a Movable Camera

Abstract

In this paper, we propose a recognition method comparing characteristics of three-dimensional CAD information with those of input images taken by a movable camera. At first, three-dimensional coordinates of vertexes of an object are determined and the existence of curved surfaces and holes are checked by input images, which are taken by a movable camera. Next, the kind and direction of the object are determined by comparing those characteristics with CAD information. In a case of that, the object can't be determined by them, a difference of CAD information is calculated. Then the object is recognized by comparing the CAD information with input image, which includes the different part of the object.