ABSTRACT

Xu, S.; He, X., and Li, X., 2020. An improved method on U-net near coastal areas. In: Zheng, C.W.; Wang, Q.; Zhan, C., and Yang, S.B. (eds.), Air-Sea Interaction and Coastal Environments of the Maritime and Polar Silk Roads.Journal of Coastal Research, Special Issue No. 99, pp. 48-53. Coconut Creek (Florida), ISSN 0749-0208.

Object recognition near coastal area is the hotspot in recent years. However, the problems such as lack of precision, low efficiency and low fault tolerance of these traditional methods are an obstacle of hyperspectral image. This paper proposes an improved method on U-net deep learning model to identify these categories. Firstly, data enhancement processing is implemented on mini-batch×mini-batch×3 patches of cropped randomly images to improve the generalization ability of training network. Secondly, an improved U-net deep learning model is developed to extract the feature information automatically. Next, the improved model can be trained repeatedly by the cross-entropy loss function and INadam optimization algorithm to seek better results. Finally, obtained optimal model is used to extract the object information accurately. Simulation and experimental results show that the proposed method has better detection performance, higher training accuracy with lower loss fluctuation than state-of-the-art methods.

This content is only available as a PDF.
You do not currently have access to this content.