Reverse Attention Dual-Stream Network for Extracting Laver Aquaculture Areas From GF-1 Remote Sensing Images

Extracting laver aquaculture areas from remote sensing images is very important for monitoring and scientific management. However, due to the large differences in spectral features of caused by factors such as different growth stages harvesting conditions, traditional machine learning deep methods face great challenges achieving accurate complete extraction raft areas. In this article, a reverse attention dual-stream network (RADNet) proposed with weak responses comprehensively considering both boundary surrounding sea background information. RADNet consists stream segmentation stream. Considering weaker certain areas, we introduce module amplify inapparent To suppress response nonboundary details stream, design module, which guided high-level semantics The structural information area learned will be fed back through specially designed guidance module. study conducted Haizhou Bay, China, verified using self-labeled GF-1 multispectral dataset. experimental results show that model performs better extracting compared SOTA models.