Hybrid-polarity for Sar Coastline Detection and Classification

Polarimetric SAR imagery can provide detailed level of information both to detect the border line and to classify an immediately close area regarding to urban areas, ports infrastructures and vegetation. Compact-pol features are derived from Radarsat-2 full-pol data and their combination is applied both to edge detection and to coast classification in Sardegna’s South Coast Italy. Initialy, coastline is extracted based on feature which presents the smallest sea backscattering response. Then, wavelets filter are achieved to reduce the noise and a clustering technique, e.g. k-means, is employed together with usual edge detection algorithm (canny algorithm). Further polarimetric features are combined according to the Principal Component Analysis (PCA) and the most significative components are selected to classify the coast according its polarimetric characteristics.