一种新的水体指数及其稳定性研究

New water index and its stability study

针对地表水提取时易受山体阴影、建筑物阴影和雪等因素影响精度问题,本文提出一种新的研究思路:将地物分为水和非水两种类型,统计各地物的反射率,研究阴影之类的低反射率地物与雪之类的高反射率地物与水体的关系.在水体和其他非水体像元最大反射率基础上,构建新的水体指数模型Water Index(WI2019).用Landsat 8 Surface Reflectance Tier 1影像的实验结果表明,与已有的7种水体指数相比,WI2019平均总体精度、Kappa系数、错分与漏分误差分别是97.98%、0.9569、3.88%和1.22%,总体上与其他水体指数比较具有一定的优势.实验发现,在有雪环境下,缨帽变换湿度指数、改进的归一化水指数、自动水体提取指数、水体指数2015和多波段水体指数均不能有效的进行水体提取.本文的创新在于利用Landsat 8的三个波段,构建了一个可以抑制阴影和雪的新的水体指数,补充了其他已有的水体指数的不足.本文方法能够稳定地提取地表水,且能够稳定地去除山体阴影、建筑物阴影和雪等的影响.

In view of the accuracy of surface water extraction,such as mountain shadow,building shadow and snow,this paper proposes a new research idea:the ground object is divided into two types,water and non-water,and the reflectivity of each object is counted.Study the relationship between low-reflectivity features such as shadows and high-reflectivity features such as snow and water.Based on the maximum reflectivity of water bodies and other non-aqueous bodies,a new water index model,Water Index(WI2019),was constructed.The experimental results of Landsat 8 Surface Reflectance Tier 1 image show that compared with the existing seven water body indexes,the average overall accuracy,Kappa coefficient,commission errors and omission errors of WI2019 are 97.98%,0.9569,3.88%and 1.22%,respectively.In general,it has certain advantages compared with other water bodies.The experiment found that in the snowy environment,tasseled cap wentness,modified normalized differential water index,automated water extraction index,water index 2015,and multi-band water index could not effectively extract the water.The innovation of this paper is to use the three bands of Landsat 8 to construct a new water body index that can suppress shadows and snow,complementing the existing water body index.The method can stably extract surface water and can stably remove the effects of mountain shadows,building shadows and snow.