面向像元-对象双尺度的不透水面遥感检测

Remote sensing detection of impermeable surface based on dual scaling with pixel-object

为了改善像元级遥感不透水面提取时存在的椒盐现象,提出了一种面向像元-对象双尺度的不透水面遥感检测方法.首先基于建筑、植被、水体等多种光谱指数,在像元尺度上提取不透水面的分布范围,然后对遥感图像进行多尺度分割,并通过设置高阈值和低阈值来处理图中的碎斑和空洞.利用双尺度不透水面检测方法对天津市精武镇多年的遥感图像进行不透水面检测,结果表明,该方法检测的正确率为80%左右,漏检率低于10%,检测精度高于光谱指数方法.

In order to decrease the salt and pepper phenomenon in extracting pixel level remote sensing impermeable surface,a pixel-object oriented dual scale remote sensing detection method for impervious surface was proposed.Impermeable surface distribution was firstly extracted by a variety of spectral indexes about building,vegetation and water on pixel scaling.Then multiresolution segmentation was carried out for remote sensing image.The spots and holes in the image were further processed through setting high and low threshold values.The dual scale impervious surface detection method was used to detect the impervious surface of the remote sensing images of Jingwu in Tianjin for many years.The results showed that the accuracy of the proposed method was about 80%,and the missing detection rate was lower than 10%,obtaining higher accuracy than the detection rate with spectral index method.