主成分分析法(PCA)在SIFT匹配算法中的应用

Application of PCA Method in SIFT Matching Algorithms

针对传统SIFT匹配算法数据量大、耗时长的问题,采用了主成分不变特征变换(PCA-SIFT)匹配算法。PCA-SIFT匹配算法将传统SIFT算法中的直方图法换做主元分析法,降低了传统SIFT特征描述符的维数,减少了数据量,提高了匹配效率。首先提取出两幅待匹配图像中的所有特征点及其特征向量,其次将提取出的特征向量采用距离比阈值筛选出匹配点对,再采用RANSAC法消除错配,最后得到精确的匹配结果。实验结果表明,PCA-SIFT算法较稳定、精确、快速。

An algorithm based on PCA-SIFT feature detection method（ Principal Components Analysis-scale invariant keypoints） is introduced into im- age feature detecting and matching for better real time performance and higher precision. Traditional SIfT method has a large amount of data,and needs long time, PCA-SIFT changed histogram method for main element analysis method, effectively reducing the dimension of the feature descriptor. The ex- tracted feature points are matched with the euclidean distance ratio, and then using the RANSAC algorithm to remove false matching. The experimental results show that the PCA-SIFT algorithm is more stable, more accurate and more rapid.