作为横跨3个国家(尼泊尔、印度、中国)的国际跨界河流——柯西河流域,地形高差巨大,土地覆被结构组成复杂,进行土地覆被的自动分类研究具有典型意义。基于面向对象方法多源遥感数据、训练规则、丰富的细节信息为复杂土地覆被自动分类研究提供了可能。选择合适的影像分割特征和最优分割尺度,按照数据挖掘中的规则顺序逐步进行各个土地覆被的提取。总体精度说明分类结果与野外点相一致的概率能达到90.05%,说明国际跨界河流土地覆被分类方法是可行的,分类结果是准确、可信的。 As an international cross-border river across the three countries (Nepal, India and China), the Cauchy River Basin, the terrain has a huge elevation difference and the composition of the land cover is complex. Therefore, it is of typical significance to study the automatic classification of land cover. Based on the object-oriented method of multi-source remote sensing data, training rules and abundant detail information, it is possible to study the automatic classification of complex land cover. Select the appropriate image segmentation features and the optimal segmentation scale, according to the rules of the data mining step by step for each of the land cover extraction. The overall accuracy shows that the probability that the classification result agrees with the field point can reach 90.05%, indicating that the classification method of land cover by international transboundary rivers is feasible and the classification result is accurate and credible.