PCA-SIFT 特征匹配算法研究

摘要

SIFT特征具有旋转、尺度和明暗变化的不变性，在图像匹配中得到广泛应用。针对SIFT 特征匹配中耗时长、匹配点对少、准确率低的问题，提出PCA‐SIFT特征匹配的方法。使用更加精简的方法构建特征点描述向量，通过预先构建的投影矩阵对描述向量进行主成分分析，降低描述向量的维度从而提高了特征匹配的速度，同时降维也对描述向量进行了去噪提纯，使得匹配更加准确。实验证明，利用PCA‐SIFT 特征进行匹配在降低匹配耗时的同时，增加了匹配点对，匹配准确率也得到提高。%SIFT is widely used in the feature matching because it is resistant to the rotation ,scale and illumination changes .The matching method of PCA‐SIFT is proposed aiming at the problems of time‐consuming ,fewer matching points and low accuracy rate in SIFT .A simplified method is adopted to build the descriptor .T he process of PCA reduces the dimensionality of the key point descriptors and improves the matching efficiency .In addition ,reducing the dimension is also helpful to denoise the description vector ,which makes feature matching more accurate .The experiments prove that the PCA‐SIFT proposed in the paper can reduce the matching time ,increase the number of matching points and improve the accuracy rate .