立体视觉中的特征点匹配算法研究

立体视觉中的特征点匹配是计算机视觉中的重要组成部分,是获取三维几何信息的重要步骤。特征点匹配算法可遵从下述处理过程:(1)特征点的选择;(2)点匹配。本文比较了经典的H arris、SU SAN和相位一致性角点检测算法,得出通过相位技术找出的边缘特征点具有良好的抗噪性和稳定性,再结合基于灰度相关及外极线约束的双向匹配技术,可获得很好的匹配效果。最后,给出了实验结果。

Study of Feature Points Matching Algorithm in Stereo Vision

The feature points matching algorithm in stereo vision is one of the important tasks in computer vision and a key step to obtain 3D information.The feature points matching algorithm can be divided into two procedures: 1.searching the feature points;2.points matching.After comparing the classical methods of Harris,SUSAN and Phase Congruency Corner detection,we find that it has good performance in presence of noise and stability by phase technique.Then combined with correlation and constraint of epipolar line method,the stereo matching algorithm has a good experimental result which has been shown in this paper.