Sentinel-2影像的云南千湖山细小冰湖提取方法

Extraction method of alpine small glacial lake in Qianhu Mountain area of Yunnan province based on Sentinel-2 image

针对复杂高山峡谷地区,冰湖监测容易受山体阴影和积雪的干扰,精度难以满足人们对冰湖信息的需求的问题,该文基于Sentinel-2影像和遥感影像分析技术分析云南千湖山片区高山冰湖,针对冰湖光谱信息复杂特性,采用不同方法探讨冰湖信息提取,并结合地形和同时期的谷歌影像矢量化结果对比分析不同方法对冰湖的提取效果。结果表明:(1)不同冰湖提取方法中,采用红边波段比值法和面向对象的方法与人机交互解译结果相比效果最好,Kappa系数达85%以上且能够准确地将山体阴影和冰湖区分开;(2)从数量和空间分布看,千湖山片区的冰湖分布在海拔4000 m左右,集中在香格里拉县小中甸乡和上江乡局部地区,以三碧海、大黑海为中心,向四周散开,千湖山片区总共约有冰湖305个,沿着沟谷分布。

Aiming at the problem that in complex alpine valleys, glacial lake monitoring is easily disturbed by mountain shadow and snow cover, and the accuracy cannot meet people’s demand for glacial lake information, based on Sentinel-2 image and 3 S technology, this paper made a comprehensive analysis of alpine glacial lakes in Qianhu Mountain area of Yunnan province. In view of the complex characteristics of spectral information of glacial lakes, different methods were used to explore the extraction of glacial lake information. Moreover, based on topography and Google image vectoring results of the same period, the research on the extraction of glacial lakes by different methods was compared and analyzed. The results showed:(1)among the different glacial lake extraction methods, the red edge ratio method and the object-oriented method had the best effect compared with the human-computer interaction interpretation results, with the accuracy of more than 85% and the ability to accurately distinguish the mountain shadow from the glacial lake.(2)From the perspective of quantity and spatial distribution,the glacial lakes in the Qianhu Mountain area were distributed at an altitude of about 4 000 meters,concentrated in the Xiaozhongdian township and Shangjiang township of Shangri-la county.Centering on the Sanbihai and the Daheihai,they were scattered around.There were about 305 glacial lakes in the Qianhu Mountain area,distributed along gullies.