改进OTSU算法的砂土CT图像分割及组构分析
Improved OTSU Algorithm for Sand CT Image Segmentation and Fabric Analysis
段欣睿 1李学丰 2樊国伟 3郭阳1
作者信息
- 1. 宁夏大学 电子与电气工程学院,宁夏银川 750021
- 2. 宁夏大学土木与水利工程学院,宁夏 银川 750021;宁夏大学 固体力学研究所,宁夏银川 750021
- 3. 宁夏大学物理学院,宁夏 银川 750021
- 折叠
摘要
针对砂土各向异性的量化这一关键问题,以CT技术获取的砂土试样断面图为研究对象,首先采用直方图均衡化和中值滤波法对图像预处理,降低图像噪声、提高分辨率,其次运用数学形态学改进的OTSU算法有效克服了非均匀光照的影响,最后利用分水岭算法优化重叠颗粒之间的粘连情况,从而提高分割性能,得到清晰的二值化图像.结合环形测试线法对砂土颗粒内部孔隙的定向分布进行量化,并判断其宏观力学性.结果表明,经该方法处理后,提取的细观参数准确,能够清晰地展现砂土颗粒细观组构.验证了砂土颗粒确有典型的各向异性,证明了该方法的有效性和可靠性.
Abstract
Sand is widely used in earth-rock dams,foundation treatment,geotechnical reinforcement and other engineering applications,and there is a close relationship between its mechanical properties and internal structural characteristics.Currently,the relationship between the macro mechanical properties and micro-properties of sand is not well understood.This paper addresses the key scientific problem of quantifying the anisotropy of sand by using the profile of sand samples obtained through CT technology as the research subject.First,histogram equalization and median filtering are applied to reduce image noise and enhance resolution.Second,OTSU algorithm improved by mathematical morphology is employed to effectively mitigate the effects of non-uniform illumination.Finally,the watershed algorithm is utilized to optimize the adhesion between overlapping particles,thereby enhancing segmentation performance and producing a clear binary image.The directional distribution of internal pores within sand particles is quantified using the ring test line method,and its macroscopic mechanical properties are assessed.The results indicate that the extracted meso-parameter information is accurate and effectively illustrates the meso-fabric of sand particles processed by this method.The typical anisotropy of sand particles is verified,which proves the effectiveness and reliability of the proposed method.
关键词
CT图像/图像处理/砂土颗粒/细观特性/孔隙分布Key words
CT images/image processing/sand particles/mesoscopic characteristics/pore distribution引用本文复制引用
基金项目
宁夏科技创新领军人才计划项目(KJT2019001)
国家自然科学基金项目(12162028)
宁夏回族自治区多尺度力学及其工程应用创新团队项目(2021)()
出版年
2024
国家科技期刊平台
NSTL
万方数据