Classification of Targets in SAR Images Using SVM and k-NN Techniques

In this paper, a method developed for classification of various military target types acquired from Synthetic Aperture Radar (SAR) images is described. For classification, first images are enhanced and segmentation is performed. Then, in the feature extraction step, the use of Modified Radial Function - (MRF) based features is proposed, which had not been used in previously for SAR-based classification studies in the literature. In addition to MRF, the mean of the segmented image and ellipse axis rate are used as features to increase the classification accuracy. A classification accuracy of 93.34% has been achieved by using Support Vector Machines (SVM) and k-Nearest Neighbors (k-NN) classifiers.