中文摘要：

为表述黄土湿陷前后的微结构特征，引入谱系聚类分析法，对采集的21张黄土扫描电镜图像的微结构参数进行研究．根据聚类原理，定义了最长距离法、最短距离法以及ward法的类间距离。给出了参数间具体的聚类过程．研究表明，对黄土湿陷前后微结构进行评估时，最长距离法和Ward法具有较好的稳定性和可重复性，最短距离法无法客观评价其特征．采用的谱系聚类法，可以综合评估域内颗粒体几何属性方面的差别．

英文摘要：

In order to visually describe the microstructure characteristics of loess wet depression, the microstructure parameters of the twenty-one SEM images of loess were studied by clustering analysis method. According to the clustering principle, the three types of inter-class distance in the longest distance method, the shortest distance method and the Ward method were defined respectively, and the specific clustering process was given. The research shows that, when the microstrueture characteristics of loess wet depression were evaluated, the longest distance method and the Ward method have good stability and repeatability, but the shortest distance method cannot objectively evaluate its characteristics. The pedigree clustering method can be used to evaluate the difference of the geometrical properties of the particles in region.