الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 174 |  |  |
| | | from | 174 | | |
Abstract
This study introduces the development, validation, and application of a new algorithm developed for the detection of soil particles in digital soil images intelligently and properly. This developed algorithm is called Soil Grain Recognition (SGR) algorithm. Digital image processing nowadays plays an essential part in many applications by providing fast and reliable methods to obtain useful data from images or videos and process them for a specific purpose. Furthermore, this study investigates the watershed algorithm for the detection of soil particles in the edge maps obtained from edge detection algorithm. The method presented here offers several improvements over existing methods by overcoming the limitations in these methods such as the proper filling of detected contours. In addition, the SGR algorithm takes lesser time in the analysis on images. A number of different samples is prepared in the lab for validation. These samples includes various types of samples, grains size and gradation. The aim of this research is to assess the degree of reliability of image analysis methods to measure grains size and draw GSD curve. The new algorithm is capable of detecting the particles included in the images efficiently and then calculate their properties such as the area, perimeter, and length of long, mean and short axes. The calculated properties are further used in the estimation of each grain size and draw the grain size distribution curve. The thesis verifies the degree of accuracy of using this developed algorithm by comparing the GSD curves obtained from image analysis algorithm and mechanical sieving in the lab. Furthermore, this research investigated using different algorithms for edge detection and recommended using Holistically Nested Edge Detection algorithm. This algorithm gives the most suitable edge map images showing the boundary of soil and rock particles on black background of the same dimensions of the input image. Comparison between mechanical sieving curve and image analysis methods shows low significance between both curves based on a variety of prepared samples in the lab given high quality images are captured. Moreover, rock fragments are analyzed to estimate their size and accordingly their mass to encourage using image analysis as a reliable method for measuring the size of large scale particles.