Planning for Earth Imaging Tasks via Grid Significance Mapping

Jinwen Li, Youmei Pan, Xiaobin Zhu, Fanjiang Xu, Peng Wang

Abstract: Most of the tasks for an Earth-imaging satellite can be converted into a series of point and region observations which need to be modeled and solved separately. For example, dense points observation is usually carried out by combining nearby points of interest through preprocessing such as temporal or spatial clustering. While the selection of this aggregated discrete observation opportunities can be solved as an optimization problem, the imaging for larger areas relies on the decomposition of regions of interest as a combination of observable strips to achieve a higher regional coverage. The lack of uniform representations for such planning problems leads to a variety of divide-and-conquer schemes which are not mutually compatible. This means researchers need to model the specific task planning problems in specific ways by sacrificing flexibility and efficiency. We propose to solve Earth-imaging task planning using a novel representation of grid which can map points and regions of interest to a uniform quantification according to significance distribution in a grid cell. A multi-satellite collaborative observations is then proposed using this grid representation to image a large number of targets by arranging dozens of satellites. The efficacy of the proposed representation and modeling is demonstrated in the experiment, showing the merits of the proposed method.