Multitemporal polarimetric SAR data fusion for land cover mapping

Abstract

With the development of the polarimetric Synthetic Aperture Radar (SAR), the research on the land cover classification using SAR data has developed rapidly. But the classification accuracy is serious affected by the speckle noises on the SAR image. A new method which combines the advantages of multi temporal SAR data and quad-polarization SAR data was presented in this paper. A method of multitemporal SAR data fusion was used to eliminate the effect of the speckle noises on the SAR image. An area of 12×17km was selected as test area. 6 multi temporal RADARSAT-2 images were used in this study to contact the land cover classification. The results show that different land cover type denote different backscattering mechanism and the backscattering coefficient value of different land cover type varies as function of time. Based on the fusion result of multi temporal polarimetric SAR data, the land cover classification was fulfilled and the results show this method can effectively distinguish the man-made buildings, forest, farm land and water. The speckle noise was obviously reduced and the visual appearance of the SAR fusion image well improved.