多尺度空间下的云南山地流域遥感土地利用分类对比研究

A COMPARATIVE STUDY OF THE REMOTE SENSING LAND USE CLASSIFICATION MODEL FOR THE MOUNTAIN WATERSHED AREA IN YUNNAN PROVINCE

以云南山地流域为研究区域,利用实证理论方法及3S(RS、GIS、GPS)技术手段,进行了多尺度空间下多模型的遥感土地利用分类对比研究。结果表明:率先建立地学主导因子下的遥感地学分异规律,对提高云南遥感土地利用/土地覆盖(LUCC)分类精度效果显著;遥感分类模型效应与尺度空间有较大的关联性;对大区域及地形因子和地貌差异大的研究区,构建专家系统模型和推理分类模型是提高遥感分类精度的有效途径。

In this study, the authors selected Jinghong City along the Meigong River and Pingbian County along the Honghe River as the study areas, where the land use classifications under the same spatial scale were compared. The results were verified with the help of ground real data and analyzed by Kappa statistics. The results show that ： （ 1 ） It is essential for Yunnan Province to establish a hierarchy classification scheme based on geomorphology characteristics ; （2） The effect of different classification schemes varies with spatial scales, e. g. , on the small scale, the classification based on analyzing statistics of spectral reflectance yields acceptable accuracies. On the larger scale, however, due to the influence of landform, the expert system or inductive classification may improve the mapping accuracy.