光学学报, 2021, 41 (2): 0201002, 网络出版: 2021-02-27
基于UNet深度学习算法的东海大型漂浮藻类遥感监测 
下载： 1202次
Remote Sensing of Floating Macroalgae Blooms in the East China Sea Based on UNet Deep Learning Model
海洋光学 大型漂浮藻类 遥感监测 深度学习 语义分割 UNet oceanic optics floating macroalgae blooms remote sensing monitoring deep learning semantic segmentation UNet
摘要
基于语义分割神经网络UNet,利用GOCI(Geostationary Ocean Color Imager)卫星传感器数据,构建出能够有效提取大型漂浮藻类的深度学习模型,实现了对大型漂浮藻类信息端到端、像素到像素的分割识别。验证结果表明:所提出的深度学习模型对验证集中大型漂浮藻类的平均识别精度达到88.54%;通过与传统的归一化植被指数法和替代型漂浮藻类指数法进行对比,发现基于UNet构建的大型漂浮藻类监测模型具有更高的准确率且受云的影响较小。利用UNet大型漂浮藻类提取模型的识别结果对2017年东海藻类暴发过程进行了分析,模型显示出很好的实用性。
Abstract
This paper proposed a deep learning model based on a semantic segmentation neural network (UNet) for extracting floating macroalgae blooms effectively from the data of Geostationary Ocean Color Imager (GOCI) satellite sensors, achieving the end-to-end and pixel-to-pixel segmentation and recognition of the information of floating macroalgae blooms. The validation results show that the average recognition accuracy of the deep learning model for floating macroalgae blooms in the validation set can reach 88.54%. Compared with existing methods for detecting floating macroalgae blooms, including normalized difference vegetation index (NDVI) and alternative floating algae index (AFAI), the constructed model based on the UNet for monitoring floating macroalgae blooms has high accuracy and is less affected by clouds. Consequently, the recognition results of the UNet based model for floating macroalgae blooms are successfully applied to analyzing the outbreak process of floating macroalgae blooms in the East China Sea in 2017. The proposed model indicates a good applicability.
李潇凡, 王胜强, 翁轩, 孙德勇, 张海龙, 焦红波, 梁涵玮. 基于UNet深度学习算法的东海大型漂浮藻类遥感监测[J]. 光学学报, 2021, 41(2): 0201002. Xiaofan Li, Shengqiang Wang, Xuan Weng, Deyong Sun, Hailong Zhang, Hongbo Jiao, Hanwei Liang. Remote Sensing of Floating Macroalgae Blooms in the East China Sea Based on UNet Deep Learning Model[J]. Acta Optica Sinica, 2021, 41(2): 0201002.
PDF全文
