中文摘要：

我国至今尚未形成一个公认的地貌分类系统,传统地貌类型的划分主要依据海拔及起伏度,较少考虑地貌的完整性原则,且分类结果琐碎。为了避免此类问题,以我国青藏高原及其邻近地区1000 m分辨率的SRTM DEM为数据源,采用面向对象思想,基于eCognition软件,利用多尺度分割、局部方差法以及决策树分类法自动划分了地貌形态。结果表明：（1）在分割尺度范围为10-1400、步长为100时,青藏高原及其邻近地区的最佳分割尺度为400;（2）依据平均高程及标准差的大小,青藏高原及其邻近地区可划分为极高山、大起伏高山、高丘、小起伏高山、大起伏中山、小起伏中山、高海拔平地、低海拔平地八种地貌类型。相比依据海拔及起伏度的划分方法,分类结果更能考虑地貌的完整性原则,且具有高效便捷性,划分结果更平滑,为我国地貌类型的划分提供了参考。

英文摘要：

Our country has not yet formed a recognized landform classification system. Traditional landform classification was based primarily on altitude and degree of relief（ ruggedness）,less consideration of the principle of the integrity of landscape,and it is trivial in detail. To avoid such problems,SRTM DEM data with a resolution of1000 m collected for the Tibetan Plateau and its adjacent areas were exemplified to introduce a updated landform division,which took advantage of the object-oriented programming and e Cognition software,and utilized multi-scale segmentation,local variance method and decision tree classification to automatically realize landform classification. Results showed that：（ 1） As segmentation scale was in the range of 10-1400 with step size 100,the optimal segmentation scale for Tibetan Plateau and its adjacent areas was 400;（ 2） According to the average elevations and standard deviations,the Tibetan Plateau and its adjacent areas can be divided into eight kinds of landforms：extremely high altitude mountain; high altitude mountains with large undulating; high hill; high altitude mountain with small undulating; intermediate altitude mountain with large undulating; intermediate altitude mountain with small undulating; high altitude plains and low altitude plains. Compared to the traditional method whose classification depends on altitude and degree of relief,this method proposed here can better reflect the integrity of the principles of landscape,and it is more efficient and convenient to defining landform with a much smoother result. It would be a good reference to Chinese landform classification.