GRAIN-SIZE DISTRIBUTION OF GRANULAR SOILS BY COMPUTER VISION

A semi-automated technique for obtaining the grain-size distribution ( GSD) of granular soils using computer vision is presented. Backlighted digital images of a soil specimen dispersed over a glass specimen pla te are acquired at three different magnifications. images of the speci men are acquired by placing the specimen plate randomly beneath the fi eld of view of a charged-coupled device (CCD) video camera The size of particles with projected areas from 50 to 2000 pix(2) is measured in each image. Multiple images are acquired at each magnification until t he measured size distribution of particles counted at that magnificati on stabilizes. Probabilistic corrections are then used to obtain a sta tistically unbiased GSD from the image data obtained at all three magn ifications. A comparison of GSD data for two uniform and two nonunifor m soils using both computer vision and sieving is presented.