Trinarization of μX-ray CT images of partially saturated sand at different water-retention states using a region growing method

Abstract

The trinarization of micro-computed tomography (CT) images for partially saturated soils at different water-retention states has been performed to clearly identify the three phases, i.e., the soil particles, the pore water and the pore air. We have proposed a trinarization technique for partially saturated soils whose histograms of the gray values for the three phases overlap each other. The segmentation method used in this study is the region growing method that ensures the spatial continuity of the phases extracted by the segmentation. Micro CT images of a dense sand specimen during the wetting process in a water retention test have been obtained. It has been found that the trinarization of the CT images in a high pore saturation regime provides reasonable results, while that in a low pore saturation regime overestimates the local void ratio. This is because the gray values of the mixels of the soil particle phase and the air phase, due to the partial volume effect, are similar to those of the water phase. It is necessary, therefore, to validate the trinarization results, by a comparison with the test results, because it is difficult to theoretically evaluate the partial volume effect. The correction of the tolerance value for the low pore saturation case with validation has provided better trinarization results. Through the trinarized CT images, the form of the existing pore water at different water-retention states has been discussed.