基于分形维数和形状因子的土体物理性质与图像特征的关系研究

发布时间：2018-04-29 09:31

本文选题：土壤 + 物理性质　；参考：《哈尔滨工业大学》2015年硕士论文

【摘要】：土壤物理特性是土壤分类的依据,也为土壤力学跟水文性质提供初步判断,但是通常情况下,获得土壤物理特性的实验方法是十分繁琐、耗时的,并且带有一定的主观性,因此,给岩土工程师和实验室人员带来了巨大的挑战。为了将实验方法的不足之处减小到误差允许的范围以内,得到更加客观的土壤特性,本文探讨了应用数字图像处理技术(DIPA)代替广泛应用的实验室方法获得土壤物理特性的可能性,数字图像处理技术是通过图像特征的分形维数和形状因子与土壤特性之间存在的联系,得到土壤的物理特性(含水量,干密度,孔隙度和孔隙比)。针对土壤中颗粒形状复杂无规律性,本文应用分形维数(FD)来描述不规则颗粒的形状,形状因子描述其几何特征。按美国材料与试验协会(ASTM)标准采集土壤样本并进行实验,得到土壤物理特性,然后将相同的样本颗粒破碎到合适的大小,拍摄数字照片,进行数字图像处理,得到图像特征。对图像特征进行数理统计分析,通过回归方程计算土壤物理特性,计算过程中,选取合适的方程模型和相关系数是得到理想结果的前提。用传统实验结果验证图像处理结果的准确性,该过程中采用不同地方的土壤样本,并将新得到的样本数据回带到模型方程,并计算模型误差。通过分析,本文得到了土壤特性与图像特征之间的曲线模型。描述土壤物理特性与分形维数之间关系时,相关系数r取值范围为0.3-0.4,描述物理特性与形状因子之间的关系时,r的取值范围为0.5-0.6。将验证样本的分形维数和平均形状系数带入模型方程,关于分形维数的模型误差为0.6%-3.57%,形状系数的模型误差为1.75%-5.13%。结果表明,本文提出的数字图像处理技术能比较准确的计算土壤物理特性,具有广阔的发展前景,同时也是一种量化土壤颗粒几何特性的有效方法。
[Abstract]:Soil physical properties are the basis of soil classification, and also provide preliminary judgment for soil mechanics and hydrological properties. But in general, the experimental method to obtain soil physical properties is very cumbersome, time-consuming and subjective. Therefore, it brings great challenges to geotechnical engineers and laboratory personnel. In order to reduce the inadequacies of the experimental method to the range of errors allowed, and to obtain more objective soil properties, In this paper, the possibility of obtaining soil physical properties by using digital image processing technique (DIPA) instead of the widely used laboratory method is discussed. Digital image processing is based on the relationship between the fractal dimension and shape factor of image features and soil properties, and the physical properties of soil (water content, dry density, porosity and porosity ratio) are obtained. In this paper, the fractal dimension (FDD) is used to describe the shape of irregular particles and the shape factor to describe their geometric characteristics. According to ASTM (American Association for Materials and testing), soil samples are collected and tested to obtain soil physical properties. Then the same sample particles are broken to the appropriate size, digital photographs are taken and digital image processing is carried out to obtain the image features. The mathematical statistical analysis of image features and the calculation of soil physical properties by regression equation are carried out. In the process of calculation, the premise of obtaining ideal results is to select appropriate equation models and correlation coefficients. The accuracy of the image processing results is verified by the traditional experimental results. Soil samples from different places are used in the process, and the newly obtained sample data are brought back to the model equation and the model errors are calculated. Through analysis, the curve model between soil characteristics and image features is obtained. When describing the relationship between soil physical properties and fractal dimension, the range of correlation coefficient r is 0.3-0.4, and the range of r is 0.5-0.6 when describing the relationship between physical properties and shape factors. The fractal dimension and the average shape coefficient of the verification sample are introduced into the model equation. The model error of the fractal dimension is 0.6 ~ 3.57, and the model error of the shape coefficient is 1.75 -5.13. The results show that the digital image processing technique presented in this paper can calculate soil physical properties accurately and has a broad prospect of development. It is also an effective method for quantifying the geometric properties of soil particles.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TU411

【相似文献】