Lightweight Deep Learning for Marine Target Detection in Sea Clutter

Efficient and reliable detection of marine targets in sea clutter remains a challenging task in radar signal processing. To address this critical challenge, we formulate the problem as a binary classification task using deep learning methodologies. This study introduces an innovative neural network architecture that directly processes raw time-frequency (TF) representations of radar signals, maximizing the retention of critical signal characteristics. Our experiments utilize the widely recognized Ice Multiparameter Imaging X-Band Radar (IPIX) dataset. The results demonstrate that our proposed method achieves a high Area Under the Curve (AUC) of 0.998313, with a network parameter count of 1.52 million, significantly lower than the parameter requirements of other methods. These findings underscore the potential of lightweight neural architectures for maritime surveillance systems.