面向对象的高分辨率影像土地利用分类

Object-oriented Land-use Information Extraction from High Resolution Images

选取许昌市QuickBird遥感影像数据,采用面向对象方法提取土地利用信息.通过重复性实验对影像进行最优尺度分割,综合利用高分辨率遥感影像的光谱、形状和纹理特征,分别建立耕地、林地、园地、建筑用地、道路交通用地、水域和城市绿地的特征集,采用模糊规则分类的方法实现土地利用的分类.结果表明该方法分类精度达到80.55%,Kappa系数为0.773,总体分类效果较好.

This paper selected QuickBird remote sensing image data of Xuchang and extracted land use information by object-oriented method.The image was segmented at an optimal scale through repetitive experiments.The spectral,shape and texture features of high-resolution remote sensing images were comprehensively utilized to establish the feature space.The classification result was obtained by fuzzy rules,including cultivated land,forest land,garden land,building land,traffic land,water area and urban green land.The result showed that the classification accuracy was 80.55%,the Kappa coefficient was 0.773,and the overall classification effect was better.