基于信息传输方式重构U型水体提取网络研究OA

Water body extraction from remote sensing images,as an efficient means of water body monitoring,has received widespread attention with the continuous development of deep learning technology in remote sensing.However,challenges such as multi-scale Water body extraction from remote sensing images,as an efficient means of water body monitoring,has received widespread attention with the continuous development of deep learning technology in remote sensing.However,challenges such as multi-scale feature extraction,visual noise interference,and the accurate identification of fuzzy boundaries remain,limiting the performance improvement of existing models.To address these challenges,this paper proposes an improved deep learning model,RASP-Unetplus,based on the classical U-Net architecture and incorporating a residual structure,empty space pyramid pooling(ASPP),and an attention mechanism.By introducing the idea of residual learning,the common problems of gradient vanishing and gradient explosion in deep networks are effectively mitigated;at the same time,the combination of ASPP and the attention mechanism not only highlights the information of important features but also significantly enlarges the sensory field of the model,which effectively solves the problems of multi-scale feature extraction and boundary ambiguity.The experimental results show that,compared with a variety of classical state-of-the-art models,RASP-Unetplus achieves the best comprehensive performance,with an intersection-over-union(IoU)of 94.92%,fully demonstrating the model's effectiveness and superiority in remote sensing imagery water body extraction tasks.