Coastal Aquaculture Area Extraction Based on Self-Attention Mechanism and Auxiliary Loss

With the development of deep learning in satellite remote sensing image segmentation, convolutional neural networks have achieved better results than traditional methods. In some full networks, number network layers usually increases to obtain features, but gradient disappearance problem occurs when deepens. Many scholars obtained multiscale features by using different calculations. We want structure while obtaining contextual information other means. This article employs self-attention mechanism and auxiliary loss (SAMALNet) solve above problems. adopt strategy atrous spatial pyramid pooling module extract considering information. add overcome problem. The experimental extracting aquaculture areas Jiaozhou Bay area Qingdao from high-resolution GF-2 images show that, general, SAMALNet achieves compared with UPS-Net, SegNet, DeepLabv3, UNet, DeepLabv3+, PSPNet structures, including recall 96.34%, precision 95.91%, F1 score 96.12%, MIoU 92.60%. boundaries structures listed above. high accuracy can provide data support for rational planning environmental protection coastal promote more usage area.