Automatic Target Recognition in SAR Using Deep Learning

The project uses many useful deep learning methods to automatically recognize targets in SAR images and relies on the well-known MSTAR dataset. In defense and surveillance, radar systems in many sectors are valuable and important because they capture high-resolution target images in any weather or lighting conditions. One of the main goal of this project is to assess how well CNN-based models can recognize military targets in this SAR images. The study looks at how effective CNNs are at identifying targets when the radar captures them from different angles and positions. Instead of using manually crafted features, the CNN learns directly from the SAR images to classify various types of military targets, even when the images have speckle noise or other distortions. To improve accuracy and flexibility, the model also uses data augmentation and normalization techniques. The results of the project execute and shows that deep learning provides a reliable way for real-time recognition and classification of targets using radar data.