基于Landsat8 OLI遥感影像冰湖边界分类叠加提取

青藏高原特殊的地理及其气候环境适合冰湖变化研究。传统的方法由于受到冻湖、冰川以及山体的阴影的影响,提取冰湖边界时会出现大量的错分和误分。本文基于NDWI(Normalized Difference Water Index,水体指数法)、DEM(数字高程模型)、RGI(Randolph Glacier Inventory)6.0全球冰川编目数据提出针对不同状态下的冰湖边界提取的分类叠加提取方法。结果表明,分类叠加提取方法能够实现未结冰、半结冰以及全结冰三种状态下的冰湖边界提取,并且能减少冰川以及山体阴影的影响,使用该方法提取三种状态下的冰湖的错分和漏分误差均小于2%且Kappa系数均大于0.81。

Automatic Extraction of Ice Lake Boundary Based on Landsat8 Remote Sensing Image

The special geography and climatic environment of the Qinghai-Tibet Plateau are suitable for the study of glacial lake changes.Due to the influence of frozen lake,glacier and mountain shadow,there are a lot of misclassification and misclassification in traditional methods.Based on NDWI(Normalized Difference Water Index),DEM(Digital Elevation Model),RGI(Randolph Glacier Inventory)6.0 global glacier catalogue data,this paper proposes a classification superposition extraction method for ice lake extraction boundary extraction under different conditions.The results show that the classification superposition extraction method can extract the boundary of ice lake in three states:non-ice,semi-ice and full-ice,and can reduce the influence of glacier and mountain shadow.Compared with the traditional ice lake boundary extraction method,it has higher accuracy.