Adaptive land classification and new class generation by unsupervised double-stage learning in Poincare sphere space for polarimetric synthetic aperture radars

Polarimetric satellite-borne synthetic aperture radar (PolSAR) is expected to provide land usage information globally and precisely. In this paper, we propose a unsupervised double-stage learning land state classification system using a self-organizing map (SOM) that utilizes ensemble variation vectors. We find that the Poincare sphere parameters representing the polarization state of scattered wave have specific features of the land state, in particular, in their ensemble variation rather than spatial variation. Experiments demonstrate that the proposed PolSAR double-stage SOM system generate new classes appropriately, resulting in successful fine land classification and/or appropriate new class generation.