A subpixel spatial-spectral feature mining for hyperspectral image classification

This paper presents a subpixel spatial-spectral feature minin approach for hyperspectral image classification. First, a re gional clustering-based spatial preprocessing (RCSPP) strat egy is introduced to identify the endmember signatures from the original image. Then, a partial unmixing model of mix ture tuned matched filtering (MTMF) is adopted to estimat the abundance maps. Finally, the morphological componen analysis (MCA) is adopted to decompose the abundance ma into different spatial morphological components, and the s moothness components are chosen for classification. The ex perimental results reveal that the obtained subpixel spatial spectral feature can lead to very good classification accura cies.