Key points are not available for this paper at this time.
Due to the influence of the imaging characteristics of the solar orbiting satellite and atmospheric conditions, the multi-spectral observation data often have the missing of phase image, which brings difficulties to the automatic processing of regional remote sensing data. However, the observation sequence contain complex and diverse information, it is difficult for the conventional method to characterize this information. To address this issue, we perform a deep feature-mapping models (DFMMs) for generating remote sensing sequence imagery. According to different missing conditions of phase image, DFMMs contain dynamic filter network (DFN-GM), sequence modeling network (SM-GM), and radiation properties network (RP-GM). DFN-GM introduces the dynamic filter (DFN) to capture the spatio-temporal characteristics of sequence imagery. SM-GM introduces the state information of sequence imagery for multiple image generation. RP-GM introduces cycle-consistent loss to constrain information from the source domain to the target domain. Our experiments, conducted utilizing unmanned aerial vehicle (UAV) and Landsat-8 datasets, demonstrate that DFMMs are remarkably effective in generating high-quality remote sensing datasets.
Jin et al. (Mon,) studied this question.