基于Canny-Harris-SIFT算法的立体匹配  

Stereo Matching Based on Canny-Harris-SIFT Algorithm

作　　者：江祥奎[1] 王曙光[1] 

机构地区：[1]西安邮电大学自动化学院,陕西西安710121

出　　处：《机床与液压》2013年第13期59-60,63,共3页Machine Tool & Hydraulics

基　　金：陕西省自然科学基础研究计划面上项目(2011JM8003)

摘　　要：针对SIFT算法存在计算量大、复杂度高的问题,提出一种基于Canny-Harris-SIFT算法的立体匹配新方法。该算法将Canny边缘检测、基于Harris角点的匹配方法和基于SIFT特征的匹配方法有机结合,采用Harris提取特征点,为每个特征点定义主方向和生成特征向量描述子,从而提高特征匹配效率。实验结果表明:该算法降低了特征匹配的复杂度,提高了算法的实时性和图像的正确匹配率。Aiming at the problems of large calculating amount and high complexity in scale invariant feature transform algorithm, a new stereo matching method was proposed based on Canny-Harris-SIFT algorithm. In the matching method, Canny edge detection was combined with the matching methods based on Harris corner and SIFT feature. The feature points of images were detected by using Harris operator, the main orientation for each feature point was calculated, the feature point descriptors were generated, so as to im- prove the efficiency of feature matching. The experimental results show that using this algorithm, the complexity of feature extraction and matching is reduced, the real-time performance of SIFT algorithm is improved and the correct image matching rate is ensured.

关 键 词：立体匹配 尺度不变特征变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]