Statistical blind classification of terrain surfaces in SAR images

Abstract

In this paper, the problem of detecting and classifying different types of terrain in Synthetic Aperture Radar, SAR, images is considered. SAR data are the result of the backscattered energy from the illuminated area, so SAR images cannot be considered as optical ones and, as a consequence, automatic feature extraction represents a difficult task. Due to the amount of data that must be processed, the proposal of simple and robust solutions is a field of interest in SAR processing. In this work, a K-means based blind approach is proposed for detecting and classifying different types of terrain surfaces. The method is tested on two different detected SAR images acquired by TerraSAR-X (GEC/SE products). Results show that the proposed method is able to detect the different types of terrain present in the images: water, arid land, forest, growing and urban areas.