Ship Detection using SAR Images for Maritime Vigilance

The identification of ships using the SAR methodology is vital for the security and surveillance of maritime domains, but it remains a complicated task because of the image’s intricacy and its small features. We introduce a Feature Extraction Algorithm that excludes non-ship neighborhoods along with a weight map to improve detection in heavily cluttered environments. We implement Feature Pyramid Networks, which enhances feature extraction and noise filtering. The guided attention together with the optimized Swish and Tanh activations improves spatial focus while detection is refined using Particle Swarm Optimization. The tests carried out in SAR datasets proved its efficiency in real-time detection of small vessels.