Characterizing localized deformation in granular media using level set method and image segmentation

The present study aims at characterizing localized grain motion (translation and rotation) through non-contact based deformation measurement mechanism using Level Set and Image Segmentation. It involves implementation of an algorithm in determining the grain shape using image segmentation, track its deformation field with time in dense grained 2-D images of sand. The traditional laboratory tests do provide a good estimate of the bulk or the macro properties of the soil when it is assumed as a continuum. It is the micro properties that play a pivotal role in governing the macro-mechanical behaviour, e.g., dilatancy behaviour of sand during shearing. Numerical tools viz. Discrete Element Modelling (DEM), Smooth Particle Hydrodynamics (SPH), etc. are often used to understand the micro behaviour of soil. On experimental front also, the need has thus been felt to estimate the localized deformation and bridge the gap of micro to macro response. A method has been developed to measure deformation profile of sand from the images taken during a few experiments in which a sand layer was sheared along a predetermined failure plane. A Perspex glass plate was placed atop the sand grains with all sides of it clipped with the bottom part and images were acquired with great care. Localized strain accumulation in soils manifests itself in the form of extensive straining which is visible in the form of shear bands in laboratory single element Plane Strain Tests. An attempt has been taken so as to quantify the number of sand grains (of 0.5-1 mm order) undergoing translation and rotation along with the degree of rotation or translation. The scheme implemented in estimating grain rotation can be implemented with slight modification for assessing the granular rotation in zones of excessive deformation.