Measurement of particle movement in granular soils using image analysis

The flowing behavior of granular soils in the form of localized deformations and shear bands is a difficult phenomenon to define explicitly in the modeling of soil-structure interface problems. However, for development of accurate numerical solutions for these problems, an estimation of particle movements is important in determining the nature of displacement fields within the granular media. Video images from direct shear tests were used to capture the movement of individual grains relative to the movement of a ribbed structural surface. Two different algorithms have been developed to determine particle displacements in an image sequence of the interface during a direct shear test. BMAD employs a block-matching algorithm using unit patterns of images to determine two-dimensional particle displacement vectors. MATCH, on the other hand, computes displacements based on centroid locations of segmented particles that are matched using a four-level filter algorithm in successive image frames. Measured hardware dependent noise was reduced during image acquisition by means of a frame averaging technique implemented in BMAD. BMAD was successfully employed to determine particle displacements in Ottawa sand images obtained during shearing on a ribbed surface. The results were verified with MATCH using the same image frames.