| || Object Hierarchy-based Supervised Characterisation of Synthetic Aperture Radar Sensor Images
Author : Rishabh, Ish;Rakshit, Subrata
Source : Defence Science Journal ; Vol:58(1) ; 2008 ; pp 159-167
Subject : 621.38 Electronics
Keywords : Object segmentation;Object characterisation;Image characterisation;Terrain-based matching;SAR imagery
Abstract : A method of supervised characterisation of synthetic aperture radar (SAR) satellite images has been discussed in which simple object shape features of satellite images have been used to classify and describe the terrain types. This scheme is based on a multilevel approach in which objects of interest are first segmented out from the image and subsequently characterised based on their shape features. Once all objects have been characterised, the entire image can be characterised. Emphasis has been laid on the hierarchical information extraction from the image which enables greater flexibility in characterising the image and is not restricted to mere classification. The paper also describes a method for giving relative importance among features, i.e., to give more weights to those features that are better than others in distinguishing between competing classes. A method of comparing two SAR sensor images based on terrain elements present in the images has also been described here.