Fast Double-Channel Aggregated Feature Transform for Matching Planetary Remote Sensing Images
Rong Huang, Genyi Wan, Yingying Zhou, Zhen Ye, Huan Xie, Yusheng Xu, Xiaohua Tong
Abstract
Matching planetary remote sensing images (PRSI) is essential for deep space exploration. Through matching optical images collected by different probes, the terrain of planets can be accurately mapped; however, PRSI lacks surface texture information and exhibits noticeable nonlinear radiation differences, e.g., illumination differences. It is not easy to achieve satisfactory results through traditional matching methods. In order to cope with the above problems, a new PRSI matching method has been proposed. It can extract low-high frequency features by building double-frequency scale space. In addition, we also use the non-maximum suppression strategy for rejecting overlapped feature points, which reduces the time consumption and improves the matching accuracy. The experimental results show that the proposed method can effectively match PRSI and be superior to comparable methods.