摘要
以湿陷性黄土的电镜扫描(SEM)和三轴CT扫描试验为基础,针对CT图像分辨率较低、难以实现土微结构精确量化的缺陷,通过对不同放大倍数的SEM图像进行图像分析,并从其中选择标准训练样本,利用训练样本对CT图像进行监督分类,从而达到定量化分析土的微结构的目的。通过比较CT图像基于自身灰度分级和基于SEM训练样本两种不同方法进行监督分类,结果表明基于SEM训练样本的CT图像监督分类,可以更好地量化监测黄土大孔隙、团粒、黏土集粒和矿物颗粒在固结剪切过程中的变化规律,从而为土的微结构研究提供了新的视角。
Because of the low resolution of CT images, it is difficult to quantize the loess microstructure accurately. Therefore, in this paper, SEM images are associated with triaxial CT images for the investigating of the microstructure of the collapsible loess. Firstly, the reasonable training samples are extracted from a large number of SEM images with different magnifications. The supervised classification of the CT images is carried out based on these training samples. For a comparison, supervised classification under gray classification of CT images is also made. The results show that： supervised classification of CT images based on SEM images exhibits better performance on the quantization monitoring of the change rule for loess microstructure. It is shown the high quality for monitoring larger pores, aggregates, clay aggregate particles and closed mineral substance of the loess sample during the triaxial shear test. It is believed that this supervised classification based on SEM images could provide a new sight for researching loess microstructure.
出处
《岩土力学》
EI
CAS
CSCD
北大核心
2012年第1期243-247,254,共6页
Rock and Soil Mechanics
基金
冻土工程国家重点实验室基金项目(No.SKLFSE200702)
关键词
黄土微结构
CT图像
SEM图像
训练样本
监督分类
loess microstructure
CT images
SEM images
training sample
supervised classification
