New Classification Technique based on Depolarised Target Detection

Inhalt:
Polarimetry has been demonstrating to be a valuable instrument for classification of scenery. Two approaches are classically attempted: a physical and statistical one. In this paper, a physical approach is presented, based on the partial target detector recently presented by the authors. The classification is achieved performing several detections of the typologies of areas to discriminate and deciding for the maximum among the obtained detection masks. Validation against L-band full polarimetric E-SAR data is provided showing good discrimination capability. Moreover, the proposed algorithm is compared with the Wishart classifier confirming significant agreement.