Assessing parameters of distribution particle size and pores of a coarse-textured soils using CBCT-scan image processing

CT scan was first invented by Hounsfield in the twentieth century in 1972. But it was soon used in engineering, agriculture, biology, physics, chemistry, etc. Recently, with advances in computed tomography at the global level, the use of different generations of X-rays on a micrometer scale to study some of the different phenomena in soil science has begun. Due to the lack of geotechnical and soil mechanics studies in many engineering projects, CT scan image processing method can be used as a suitable method for extracting soil particle size and other soil characteristics. The main purpose of this study: a) The use of CBCT-scan in soil science for the first time in Iran. B) Comparing the ability of CBCT-scan in terms of quality of results with conventional methods. C) Identify the best filter and binary method (threshold). Another goal of this research is to acquaint more researchers with the application of computed tomography (CT-scan) technology in soil science studies.

The sampling area for this study was located in Diwandareh-Saqez axis in Kurdistan province, where six disturbed and undisturbed soil samples were collected in a sandy area (12 samples in total). In disturbed samples, particle size distribution was measured by ASTM D421 method, and the porosity of the samples was measured directly using the fuzzy equations in soil mechanics. In a radiology laboratory, three-dimensional images of intact soil samples were taken using a Planmeca Promax 3D CBCT CT scanner. In this study, ImageJ software was used to process CBCT-scan images. With this software, the percentage of phases, number of particles and particle size can be calculated. One of the most important steps in image processing is generating binary images. A total of 17 global thresholding methods have been proposed for generating binary images in ImageJ software. In this study, 15 standard methods for generating binary images were examined and the best method was selected. The total pore volume and soil particle size distribution of each sample calculated by quantifying X-ray images were compared with the total pore volume and soil particle size distribution obtained in the soil science laboratory and performance of the CT scan method evaluated by statistical parameters including The results of the accuracy evaluation for the correlation coefficient, mean absolute value of deviations, mean square error, root mean square error, and mean absolute error percentage.

The most significant point in image processing is the image thresholding method. In this study, due to the nature of CBCT-scan images, global thresholding was preferred. From the results of image processing, it can be understood that the results of binary images with Otsu and Intermodes methods are in complete agreement with the laboratory sample. The average of total porosity of the processing image slides is 44.03%, which is approximately consistent with the calculated 45/6% for the laboratory sample. Also, the average of ineffective porosity of the samples is about 6.53%. Therefore, it can be said that the effective porosity of the samples is about 37.5%. The results of the accuracy evaluation for the correlation coefficient, mean absolute value of deviations, mean square error, root mean square error, and mean absolute error percentage were 0.98, 1.082, 1.229, 1.108 and 2.334 respectively, indicating that the use of CBCT-scan images and image processing technique can identify and evaluate the geometric properties of granular soils with acceptable accuracy. The advantages of the computed tomography method of the soil are: (1) Obtaining information from the three-dimensional structure of the soil with appropriate accuracy in a short time, (2) Non-destructiveness of this method, and (3) Accurate separation into soil phases in different energy radiations.

Using the processes defined by the authors for image processing, this technique is well able to determine some engineering features such as particle size distribution, total porosity, effective porosity and ineffective porosity. Also, the best thresholding method for binary images and processing in ImageJ is the Ostu and Intermodes method. The accuracy of the device used in this research is 0.2 mm, in other words, spaces or grains smaller than this value cannot be identified; For this reason, in the present study, the term coarse-textured soils, which means gravel to coarse-grained sand, has been emphasized. The results of evaluating the statistical parameters testify to the accuracy and ability of this method.