IMAGE-BASED ESTIMATION OF SOIL PARTICLE SIZE AS A SUPPLEMENT TO SIEVE ANALYSIS

Abstract

Particle size distribution is an important parameter influencing soil behavior. This parameter is identified using a combination of tests such as the sieve and hydrometer analyses to identify the potential characteristics and behavior of a soil sample. This study investigates the use of image processing techniques as a supplementary tool to traditional methods, enabling the rapid and detailed generation of particle size distribution curves. Twenty different soil samples were collected and prepared for imaging following controlled conditions. Only particles retained up to a No. 40 (0.420 mm) sieve were considered in the study. Camera and measurement calibrations, thresholding, and edge detection were performed to extract particle dimensions. Particle boundaries were identified to estimate particle sizes using pixel to real-world measurement conversions. At least one of the dimensions taken from the particle is expected to be of greater value than a reference value from the sieve openings where it was retained to ensure that there is an accurate result in the particle size estimation. The test results showed a close correlation between the estimated particle size distribution and sieve analysis using the maximum Feret diameter. However, there is an observable decline in accuracy when the particle size decreases. The proposed method has potential to capture particle size and be further studied for use in the development of particle size distribution curves, confirming its viability as a supplementary technique.