Knowledge-based 3-D Vision System incorporating Geometric Modeler as Deep Knowledge

点击翻译按钮获取中文摘要

In order to construct advanced image understanding system, it is important to achieve the efficient use of geometric knowledge in the knowledge-based system scheme. In particular, the efficient use of solid modeler Is important if the target objects are artificial ones. We have developed a knowledge-based 3-D vision system incorporating a solid modeler and geometric reasoning.The solid modeler allows the precise and reliable understanding of possible objects. Furthermore, it can be used to synthesize virtual images based on the understanding result. A geometric reasoning method based on the perspective geometry is employed to determine the position of pre-defined shapes in 3-D space from a single-view image. Different from other 3-D vision systems using geometric reasoning, no assumption has been made beforehand about the position and orientation of the possible object; therefore, there are six degrees of freedom.The synthesized image from solid modeler is effectively used to verify the hypothesis computed from the edge information extracted the observed image. To lesson the inefficiency of using the mathematical geometric method in the whole understanding process, we have introduced a plane recognition process in an intermediate stage.