Multi-source remote sensing images matching based on nonlinear scale space
Li Pen · Cehui kexue · 2015
Aiming at the problem that Gaussian blurring may not respect the natural boundaries of objects and reduce the matching accuracy in extracting feature points by Classic SIFT algorithm for RS image matching,the paper proposed the latest KAZE algorithm based on nonlinear and space constraint:the geometric transformation model was solved by preferred feature points and feature matching,and then the searching space of the matching points was restricted to improve the speed and accuracy of the matching,finally,the iteration method of Root Mean Square Error was used to exclude the mismatched points.Experimental results showed that the feature point extraction by KAZE would have higher stability and be easier to exclude mismatched points than that by SIFT,moreover,the effect of space constraints strategy would be better than that of traditional method,and particularly,for detail and texture blurred images,KAZE algorithm could have unique advantages compared to the SIFT.