中文摘要：

在监督TS-MRF（tree-structuredMarkovrandomfield）分割中，人工指定遥感影像的分层结构交互复杂且有一定的随意性。为了解决这个问题，提出一种新的基于集合划分的分层结构自动提取算法。该算法使用二叉树结构表示分层结构，并根据集合划分准则对遥感影像中的基本类别集合逐层划分，从而自顶向下地逐步获取分层结构。实验结果表明，该算法需要人工交互少、容易解译，且能保证监督TS-MRF影像分割的准确率和效率。

英文摘要：

It is a rather complicate and random process to manually set the hierarchical structure for the supervised segmentation based on the TS-MRF （tree-structured Markov random field） model. In order to overcome such a problem, a novel automatic hierarchical structure extraction algorithm is proposed based on the set partitioning. A binary tree is employed to describe the hierarchical structure, which is obtained by the gradually partition of the basic class set according to the partition criteria from the top to the bottom. Experiments show that the proposed algorithm needs less human interaction, can be easily interpretation, and has the ability to guarantee the accuracy and efficiency of the supervised TS-MRF segmentation.