基于多源数据的CART决策树冰川提取

CART Decision Tree Glaciers Extraction Based on Multi-source Data

介绍了一种快速简便的基于多源数据的CART决策树提取方法。以各拉丹冬冰川为例,利用TM影像提取冰川,通过与手动勾绘、最大似然法的对比,CART决策树方法具有一定的优势。从提取结果上看,CART决策树、手动勾绘、最大似然法提取的面积分别为856 km^2、858 km^2、866 km^2。以手动勾绘为标准,最大似然法提取和其他两种方法有着较大差距,CART决策树方法和手动勾绘法相差较少。总的来说,CART决策树在保证精度的同时,比最大似然法精确,比手动勾绘法简单快捷。

For better extraction of the glacier information, a kind of quick and easy CART decision tree extraction method based on the multi-source data was introduced in this paper. For example, in Geladandong Glacier, the TM image was used to extract the glacier, and by comparing it with the manual drawing and maximum likelihood method, the CART decision tree method had certain advantages. The extraction results show that the extraction areas of the CART decision tree method, manual drawing method and maximum likelihood method were 856 km^2, 858 km^2 and 866 km^2 respectively. On the basis of the manual drawing method, there was a larger gap between the maximum likelihood method and the other two methods, but relatively smaller between the CART decision tree method and the manual drawing method. In general, the CART decision tree method can guarantee sufficient accuracy and be more accurate than the maximum likelihood method while simpler and faster than the manual drawing method.