立体匹配是目前计算机视觉领域中最活跃的研究主题之一,在三维重构、对象识别与分类、图像对齐等应用中,立体匹配都是一个关键步骤.为了能够更精确地对特征点进行匹配,给出了一个新的匹配约束条件:匹配点距离相关性约束,据此提出了一种新的立体匹配算法.该算法独立于特征点的检测算法,首先使用灰度相关法进行初始匹配,然后利用最小中值平方法获得部分正确匹配点集,在此基础上加入匹配点距离相关性约束进行引导匹配,得到最后的匹配点集.实验表明,该算法能够很好地实现特征点的正确匹配.具有很高的使用价值.%Stereo matching is still the most active research theme in the field of computer vision, and it is a key step in the applications of 3d reconstruction, object recognition and classification, image alignment and so on. In order to do more precise to the stereo matching of feature points, gave a new matching constraint condition: distance correlation of matching points constraint and proposed a new stereo matching algorithm. The algorithm is independent of feature point detection algorithm, firstly, use the gray correlation method to do initial matching and then use LMedS method to get partial correct matching points set, and on the basis,add distance correlation of matching points constraint to guide the matching. Experiments show that the algorithm can achieve good matches of feature points, and has a very high value.
展开▼
