FKPCA-SIF算法在图像匹配中的应用.docVIP

FKPCA-SIFT算法在图像匹配中的应用 张宁丽1，马燕1，李顺宝2，徐衍鲁1，程玮2 (1.上海师范大学信息与机电工程学院,上海 20023;2.上海师范大学数理学院，上海 200234) 摘 要：SIFT是目前广泛应用于目标识别和图像匹配领域的算法，但其在使用过程中存在描述子维数过大，耗时时间长的缺点。针对这个问题，常用的解决办法是利用PCA算法对描述子进行降维，由于PCA是一种线性降维算法，因此它的使用具有局限性。对此，利用模糊K均值算法对其进行改进(称为FKPCA)，并用改进的RANSAC算法消除误匹配点。实验结果表明，PCA-SIFT算法和FKPCA-SIFT都很好的保持了SIFT算法原有的优点，具有很高的匹配正确率。但相对于PCA-SIFT算法，FKPCA-SIFT不仅适用于线性降维也适用于非线性降维，具有更好的匹配精度，拓展了PCA-SIFT算法的适用范围。 关键字：SIFT；PCA-SIFT；FKPCA-SIFT；RANSAC；线性降维；非线性降维 【中图分类号】TP391.4 【文献标志码 】A The Application of FKPCA-SIFT algorithm in Image Matching ZHANG Ning-li1, MA Yan1, LI Shun-bao2, XU Yan-lu1，CHENG Wei2 (1.Department of Information Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai200234, China 2.Department of Mathematic Science, Shanghai Normal University, Shanghai200234, China) Abstract: SIFT algorithm is widely used in the field of object recognition and image matching. But it also has disadvantages that sub-dimension is too large and time-consuming. For this problem, the common solution is using PCA algorithm to reduce dimension of the descriptors. But PCA is a linear dimensionality reduction algorithm, so its use is limited. For this situation, using the fuzzy K-means algorithm to improve it (called FKPCA) and eliminate false matching points with an improved RANSAC algorithm. The experimental results were shown that both of the PCA-SIFT and FKPCA-SIFT algorithm can perfectly keep the original advantages of SIFT that have a high matching accuracy. But compared with the PCA-SIFT algorithm, FKPCA-SIFT can not only be used to the linear dimension reduction, it can also be applied to non-linear situation. The FKPCA-SIFT has better matching accuracy and expands the application scope of PCA-SIFT. Keywords: SIFT; PCA-SIFT; FKPCA-SIFT; improved RANSAC; linear dimensionality reduction; non-linear dimensionality reduction 1 引言 David G.Lowe在2004年总结提出的一种基于尺度空间的特征匹配算法—SIFT，它对于尺度变换，光照和旋转都有很强的稳定性，因此广泛的应用于目标识别和图像匹配等众多领域[1]。但是它也存在维数过大、耗时过长的特点，传统的处理方法是引入PCA（主成分分析）算法，