一种改进的SIFT特征反向匹配算法

An Improved SIFT Features Reversed Matching Algorithm

对SIFT算法进行研究,根据目前在倒车辅助系统中立体匹配技术的应用,对该算法进行改进,为了解决高维空间搜索问题,在获得的特征向量匹配对上加入外极线约束条件,将搜索空间从二维图像降为一维的线性搜索,缩短匹配时间,降低了计算的复杂度,同时针对阈值增大,匹配数目增多,同时错误率增高的情况,对特征点进行反向匹配,进一步提高图像匹配的准确率。

Study the sift algorithm,and according to the appliance of stereo matching technology on vehicle＇s back-up aid system,make improvement on it.In order to solve the search problem in high dimensional space,add the epipolar constraint on the obtained feature vector matches,which can decrease the searching space from Two-dimensional image to One-dimensional linear search,then shorten the matching time,and reduce the computational complexity.At the same time,for the situation of the number of matches increasing with the threshold,and getting higher error rate,reversing matching is done,which can get further improve on the accuracy of image matching.