Grain Portraits: Quantifying Heterogeneity of Aggregate Layers through Image Analysis

An approach is presented for quantifying surface heterogeneity of granular layers based on digital analysis of images taken using low-cost, widely available equipment. The method enables rapid survey of large surface areas with high spatial discretization. Inputs to the analysis routine are RGB pictures of the aggregate surface, while key outputs are approximate size distributions and a descriptor termed "median feature size" (MFS). The routine is comprised of four main stages: contrast enhancement, noise reduction, segmentation, and data analysis/visualization. The paper discusses the application of the method for characterizing aggregate layers built at four sites. MFS data effectively capture spatial variability of the aggregate surface coarseness, providing a means to quantify material heterogeneity produced by segregation processes. Statistical analyses demonstrate a correlation between MFS and in situ hydraulic conductivity, and support the notion that the MFS is an indicator of local heterogeneities driving variability of hydraulic conductivity at the local scale.