AUTOMATIC TARGET RECOGNITION FOR SYNTHETIC APERTURE RADAR DATA

Abstract

Automatic Target Recognition (ATR) in images generated from Synthetic Aperture Radar (SAR) has become a significant focus of research in contemporary society and represents a crucial avenue of inquiry within the realm of image processing. This study presents a thorough investigation of ATR techniques applied to SAR data. The widely used MSTAR dataset is utilized for evaluating the proposed methodologies. The initial stage of the study involves feature extraction techniques which aim to capture the relevant information from SAR data and reduce its dimensionality. The extracted features are used as inputs for various classifiers including Support Vector Machine (SVM). The performance of these classifiers is compared and evaluated based on their classification accuracy. To address the issue of speckle noise inherent in SAR imagery, mean and median filters are applied as preprocessing steps before feature extraction to investigate how noise reduction techniques affect the recognition accuracy of the ATR system. The expected outcomes of this research are twofold. First, it aims to determine the most effective feature extraction method and classifier combination for SAR-based ATR tasks. Second, it intends to assess the influence of noise reduction techniques on classification performance, providing insights into the trade-off between noise reduction and classification efficiency. By comprehensively analyzing feature extraction methods, classifiers, and the impact of noise reduction techniques, this study contributes to advancing the field of ATR for SAR data. The findings will aid researchers and practitioners in selecting suitable methodologies for SAR-based target recognition, ultimately enhancing the capabilities of SAR systems in various applications