A Semantic Cognition Enhancment Network for Interference Detection in Sentinel-1 SAR Image

The existence of radio frequency interference (RFI) can cause partial image degradation, image interpretation errors, and parameter extraction deviations. The precise detection and localization of the interference is the premise step for successful mitigation. However, the shape of weak interference is changeable and lacks a unified mathematical model, while its energy difference with the surrounding background is very small, which makes it difficult to be detected in complex environment. This paper proposes a semantic cognitive enhancement network for RFI detection in Sentinel-1 SAR image. Through the fusion of dilated convolution, cross-layer connection and self-attention mechanism, it can effectively improve the detection performance of weak interference under the condition of small sample training.