在多源多光谱遥感图像中,针对匹配图像的像素之间非线性变化而导致正确匹配点对下降的情况,提出了一种基于主成分分析的多源多光谱遥感图像特征点提取算法。利用尺度不变特征变换(SIFT)算法的基本原理,首先对两幅多源的多光谱遥感图像进行主成分变换,再用变换后各自的第一主分量图像作为待匹配图像;其次,在构建尺度空间时提高尺度参数并且在进行特征匹配时,利用尺度限制条件进行匹配,这样既能提高匹配精度又能提高运算速度;最后,采用随机抽样一致性算法剔除误匹配点。这种算法能减少多源多光谱遥感图像之间像素灰度值的非线性变化对特征点匹配的影响,提取到一定数量的正确匹配点对。通过实验对比分析,所提算法比通用算法有更高的精度和更好的适用性。

It is difficult to extract the characteristics of matching points because the image intensity in multi-source images is non-linear with respect to pixels. A multi-source remote sensing image feature point extraction algorithm based on the principal oriented component analysis is proposed. In the proposed algorithm, a principal analysis is first made for the multi-source remote sensing images, Secondly, a scale space is built with the first principal components of the multi-source remote sensing images; and the scale parameter is increased to improve the matching accuracy as well as the operation speed of the matching of feature points. This algorithm can extract a certain number of the correct matching point pairs. Comparative analysis of experiments shows that the proposed algorithm enjoys higher precision and better stability than the general algorithms.