双摄象头三维视觉系统

BINOCULAR 3-D VISION SYSTEM

本文介绍了一种较实用的双摄象头三维视觉系统.主要讨论了视场内具有六个自由度物体的定位问题.系统以Frei和Chen方法抽取边缘,并运用了非均匀量化和预加重技术.采用了改进的Moravec兴趣算子法,实现了特征点的自动抽取.基于边缘的序贯分层配准法大大缩短了配准时间.在用三维数据获取物体姿态时.根据刚体运动总结的规则解决了工作特征点和模型点匹配时的组合爆炸问题和多义性问题.

A useful binocular 3-D vision system is introduced in this paper, the pose of an object with 6 degreesof freedom is discussed. Frei and Chen method is used to extract edges with the help of non-equalquantization and preemphasis techniqucs. Modified Moravec's interest operator method is applied to en-sure automatic extraction of feature points. Edge-based sequential registration and hicrarchical matchingsignificantly speed up the matching process. 4 rules summarized from rigid body movement is used in theperiod of locating object pose from 3-D data to solve combinational explosion and ambiguity problemsduring the matching process of workpiece feature points and model points.