从二维图象中恢复距离信息的体视匹配方法  

A Stereo Correspondence Approach of Recovering Range Information from 2D Images

作　　者：丁震[1] 胡钟山[1] 唐振民 杨静宇 

机构地区：[1]南京理工大学信息学院

出　　处：《南京理工大学学报》1996年第5期449-452,共4页Journal of Nanjing University of Science and Technology

基　　金：国防科技预研行业基金

摘　　要：利用双目立体视觉技术从二维图象中恢复三维距离信息是立体视觉研究中的一个重要研究方向。在双目立体视觉系统中，立体匹配部分是其核心部分。该文提出了一种基于边缘特征的立体匹配方法。该方法用边缘象素作为匹配基元，利用边缘的位置、差分强度和梯度方向属性，在核心线约束和视差连续性约束下，通过迭代运算实现了立体匹配，获得了较好的匹配效果，通过对匹配结果进行三维重建表明。It is an important investigative field in stereosis research that the range information is recovered from 2D images by using the binocular stereo vision. In a binocular stereo vision system, the stereo correspondence is a key part. A stereo matching approach based on edge feature is proposed in the paper. Under epipolar constraint and disparity continuity constraint, using the attributes of location and differential intensity as well as gradient orientation on edges, the procedure of the stereo correspondence is fulfilled by the iterative operation. The 3D reconstruction for the correspondence result shows that the approach is efficient and feasible.

关 键 词：立体视觉 图象恢复 距离信息 匹配 机器视觉 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN919.8[自动化与计算机技术—计算机科学与技术]