基于面向对象技术的黄土丘陵沟壑区切沟遥感提取方法研究

Study on Recognition of the Gully in Loess Hilly-Gully Region Based on Object-Oriented Technology

基于高分辨率遥感影像,提出了结合高分辨率影像的光谱、地形、几何形态和GLCM纹理信息等特征的切沟半自动面向对象提取方法,建立了一组沿径流方向计算纹理特征空间对比度和相关性的公式。以黄土丘陵沟壑第三副区甘肃天水桥子沟小流域World View-2影像数据为例,分别建立了耕地(山坡地、梯田)、果园、林地、农路、切沟的分类规则和算法,以影像的目视解译结果结合实地调查进行精度评价,分类结果显示,总体分类精度为75.17%,总Kappa系数为0.64,切沟的生产者精度为80%,用户精度为70.59%,取得了令人满意的结果。

Based on high resolution spatial image,this paper proposed a semi-automatic objected-based classification method to extract gully features using a combination of topographic,spectral,shape（geometric） and contextual information obtained from World View-2 data and a set of GLCMs metrics was calculated based on the flow direction.Taking the third auxiliary district of the loess hilly-gully area of Qiaozigou watershed of Tianshui in Gansu as a case,a rule-set was developed and tested on terrace,orchard,forest,road and gully.The classification results were evaluated by visual interpretation and field investigation which had promising accuracies,and the overall classification accuracy was 75.17%,the overall Kappa coefficient was 0.64,the producer accuracy of gully reached to 80% and the user accuracy was 70.59%.