Modified genetic algorithm-based sub-pixel mapping

Abstract

Sub-pixel mapping (SPM) is a promising way to predict the location of endmembers within mixed pixels at the sub pixel level. SPSAM is a real time processing algorithm for SPM, but it will generate a lot of isolated sub pixels because land cover classes were assigned to sub pixels in order. In this letter, a novel method is proposed to realize SPM. It contains two main steps: sub-pixel/pixel spatial attraction model (SPSAM) is first used to generate the initial SPM result; and then modified genetic algorithm (MGA) is applied as the post-processing method to obtain more accurate results. Three different sets of data are used for experiments: simple artificial images and two sets of real remote sensing images. The results show MGA/SPSAM outperforms the conventional SPM methods.