Automatic Target Recognition of SAR Images Using Radial Features and SVM

The armed forces use a variety of sensor information to locate and target enemy forces. Because of the large area and sparse population, the surveillance becomes a difficult problem. With technological advances, the armed forces can rely upon different types of image data like, infrared data, and radar data. Due to the enormous amount of data, it becomes very difficult to analyse the data without pre-processing and later to detect or classify target data. To full fill this gap this paper reports an algorithm for automatic target recognition in the battle field. This work focuses on synthetic aperture radar (SAR) images for recognizing enemy targets with more accuracy. The available images (MSTAR public open database which is freely available in open literature) is used for experimentation and training the SVM (Support vector Machine). The data will be pre-process first. The pre-processing is required to distinguish the target from clutters like building, trees etc., and non-target objects such as confuse vehicles etc., which is very much required for identifying the targets like Battle tank or armoured personnel carrier effectively. The clutters create noise which is to be removed in preprocessing. The algorithm will help to recognize specific class of the targets e.g., T-72 tank on the basis of target signature. The automatic target recognition is based on the location and orientation. The SVM is used for the classification. SVM are trained and validated with a test set to determine the best performance. The resulting SVM has a recognition rate of 98%.