DeepUNet: A Deep Fully Convolutional Network for Pixel-Level Sea-Land Segmentation

TLDR

Semantic segmentation is a key task in optical remote sensing, yet sea‑land segmentation remains challenging due to complex maritime environments and limited CNN‑based studies that could still be improved. The study introduces DeepUNet, a novel CNN for sea‑land segmentation, and evaluates it on a newly constructed challenging dataset against U‑Net, SegNet, and SeNet. DeepUNet extends U‑Net by replacing convolution layers in the contracting path with DownBlocks and using UpBlocks in the expansive path, adding U‑connection and Plus connection to enhance feature fusion. Experimental results demonstrate that DeepUNet improves accuracy by 1–2 % over U‑Net, SegNet, and SeNet, yielding more precise sea‑land segmentation in high‑resolution imagery.

Abstract

Semantic segmentation is a fundamental research in optical remote sensing image processing. Because of the complex maritime environment, the sea-land segmentation is a challenging task. Although the neural network has achieved excellent performance in semantic segmentation in the last years, there were a few of works using CNN for sea-land segmentation and the results could be further improved. This paper proposes a novel deep convolution neural network named DeepUNet. Like the U-Net, its structure has a contracting path and an expansive path to get high-resolution optical output. But differently, the DeepUNet uses DownBlocks instead of convolution layers in the contracting path and uses UpBlock in the expansive path. The two novel blocks bring two new connections that are U-connection and Plus connection. They are promoted to get more precise segmentation results. To verify the network architecture, we construct a new challenging sea-land dataset and compare the DeepUNet on it with the U-Net, SegNet, and SeNet. Experimental results show that DeepUNet can improve 1-2% accuracy performance compared with other architectures, especially in high-resolution optical remote sensing imagery.