Soft classification of hyperspectral images via multi-label learning

Nowadays, Hyperspectral image interpretation remains a challenging topic of research. Accordingly, divergent strategies have been proposed to address classification issues. Significant attention has been given, routinely, to single-label approaches. Nevertheless, few efforts have been dedicated to classify hyperspectral images from a soft classification view. In this paper, we propose a multi-label approach to perform a new classification schema considering the mixed pixel phenomena. The proposed framework includes a ranking approach to learn the most credible eighbor's labels for a weighted KNN-based model. This allows us, not only to perform classification tasks, but also to exploit the interdependence between classes. Our experimental results, conducted using a mixture of complementary features and hyperspectral datasets, demonstrate the effectiveness of the proposed framework without significantly increasing computational complexity.