Recent Advances in Deep-Learning Based SAR Image Target Detection and Recognition

Aim:

            To develop a lightweight, accurate, and high-performance YOLO-v11n model for detecting and classifying multi-class targets—ships, aircraft, oil spills, oil tanks, and military vehicles—from SAR and aerial images in real-time with low computational complexity.

Abstract:

         This project introduces a lightweight and highly efficient detection framework based on YOLO-v11n for multi-class SAR target detection. SAR images exhibit speckle noise, complex backgrounds, and multi-scale objects, making traditional detection models inefficient and unreliable. The proposed system integrates YOLO-v11n with optimized feature fusion techniques, lightweight backbone modules, and noise-resilient preprocessing to enhance detection performance in maritime and military surveillance scenarios. The system is capable of detecting ships, aircraft, oil tanks, oil spill areas, and military vehicles with high precision and robustness under complex imaging conditions. Evaluation results show improvements in mean Average Precision (mAP), reduced computational load, and improved inference speed, making the proposed model suitable for real-time deployment on edge devices and defense monitoring platforms.

Proposed System:

           The proposed system introduces an advanced multi-target detection framework built on YOLO-v11n, specifically optimized for SAR and aerial imagery. Unlike traditional or heavy CNN-based models, this system integrates a lightweight YOLO-inspired backbone, high-resolution feature fusion layers, and improved detection heads to ensure accurate identification of multi-scale objects such as ships, aircraft, oil tanks, and military vehicles. The model incorporates SAR-specific preprocessing pipelines that suppress speckle noise and enhance structural features, enabling robust performance in challenging conditions. By leveraging efficient architecture design and optimized training strategies, the proposed system significantly reduces computational cost while increasing detection accuracy and inference speed. The lightweight nature of YOLO-v11n ensures smooth deployment across edge devices, remote surveillance nodes, drones, and satellite-based platforms, making it suitable for real-time defense and environmental monitoring applications.

Advantage:

         High accuracy for multi-scale SAR targets; Faster inference with reduced model size; Robust detection under SAR noise, clutter, and low contrast; Suitable for UAVs, satellites, and edge devices; Lower false-alarm rate; Multi-class capability; Adaptable for day/night and all-weather surveillance