AUTOMATIC REGISTRATION OF CCD IMAGES AND INFRARED IMAGES OF HJ-1 BASED ON INVARIANT FEATURE AND MUTUAL-INFORMATION

摘要

With the great progress of obtaining multi-source remote sensing image from different sensors with different resolutions, a single kind of remote sensing image could not meet the requirement of applications in many fields and it needs various kinds of remote sensing images usually to be used together. So, the algorithm of registering multi-source remote sensing images has become a hot issue in today's research. Environmental satellite-1 (HJ-1) includes a B Satellite-an optical satellite, providing information for the changing trends of the disasters and ecological environment within visible bands(30m) and infrared spectral bands(150m and 300m). The great differences in gray and resolution resulting form different imaging mechanism of IR and visible images make the fact that still no one method could effectively realize higher accurate registration for both IR and visible images. Through analyzing various remote sensing image registration algorithms, this paper proposes a registration approach based on SIFT and Mutual-Information(MI) according to the imaging characteristics of HJ-1 IR and visible images. First, the improved SIFT algorithm realizes the uniform distribution of feature points, and also speeds up the extraction rate of SIFT feature points. Then, the feature points are matched by using the MI method, which greatly improves the matching accuracy. Finally, some mismatched points are eliminated by using consistency checking, which further improved matching accuracy. Many experiments on HJ-1 visible images and different resolution IR images show that the algorithm proposed in this paper can quickly achieve accuracy registration between HJ-1 CCD images and IR images. So the algorithm has great practical values.