Maritime Radar Small Target Detection Based on Echo Embedding and Self-Attention

Radar small target detection is capable of effectively identifying small targets present on the sea surface. In recent years, deep learning methods have demonstrated strong performance in various fields, such as natural language processing and image analysis. This paper proposes a method for detecting small targets on the sea surface based on echo embedding and self-attention. Firstly, the radar time-frequency feature maps are embedded into high-dimensional space using patches to enhance the network of feature representation capabilities. Then, important features are extracted locally and globally through the feature self-attention module within specific windows. Finally, the detection of maritime small targets is achieved through a linear layer. Our method is validated based on the IPIX dataset. The experimental results indicate that our method achieved promising detection performance on the four subsets of the IPIX dataset.