A design of a 3-D vision system based on geometric knowledge

Abstract

A design of a three-dimensional vision system based on geometric knowledge is described. The advantage of this model-based approach is to provide a precise understanding and strong robustness against partial occlusions for the object represented by a geometric solid model. The main reason for difficulty in recognizing partially occluded objects is that there is no object model with every variety of partial occlusion. It is a problem to obtain a customized object representation in three-dimensional space. The proposed system can synthesize various images dynamically from the geometric model description by handling the solid modeler. The depth information is extracted from a single view with geometric knowledge. the configuration of the target object is treated as a hypothesis. The generate-and-test strategy is used for hypothesis verification. >