Shape Characterization of Granular Particles using Image Based Techniques

Abstract

Granular soils with different sizes and shapes are often used in many civil engineering structures. In different contexts, several researchers have emphasized that shape of particles play a pivotal role in influencing several engineering properties such as maximum and minimum packing densities, shear strength, permeability and compressibility. However, the complexities involved in obtaining the geometrical parameters necessary to adequately compute particle shape have hampered the clear understanding of the contribution of particle shape to such properties. Researchers have attempted to characterize the shape of the particles by many conventional and advanced image based methods in the past. However, these methods suffer from many criticisms; conventional methods of shape characterization include ocular inspection of particles based on visual reference charts, which are more prone to user dependent interpretations. The recently developed image based methods deviate from the conventional and most well accepted definitions formulated by researchers in the past due to the difficulties involved in automating them. The aim of this thesis is to address this shortcoming by developing a robust methodology for accurate and precise determination of particle shape in accordance with the most widely accepted formulae in literature, which can replace the existing methods based on manual measurements, approximate visual charts and non-robust imaging techniques. For this purpose, several computational algorithms are written and implemented in MATLAB and operations are performed on particle images. These methods are developed to precisely characterize the particles shape parameters observed at three levels of scales, which are adequate for complete shape characterization. According to Barrett (1980) the particle shape features can be observed independently at three different scales, viz. macro-scale, meso-scale and micro-scale, the shape parameters such as form, roundness and surface texture falls into these three scales respectively. The macro-scale component of form (sphericity) is quantified as per the formula used in the visual chart proposed by Krumbein & Sloss (1951). In light of its continuing popularity and wide usage, the roundness concept proposed by Wadell (1932) is chosen to be the appropriate parameter for meso-scale shape representation. The micro-scale component of surface texture or roughness is measured by the conventional and widely used root mean square definition, by incorporating the use of digital filtering techniques. The distinct concept of angularity as proposed by Lees (1964) is used for effective shape representation of crushed particles. Kinematic behaviour of particles such as sliding, rolling and interlocking are dependent on the geometrical features observed at meso-scale present along their boundaries, which consequently govern the material strength and deformation characteristics. Based on precise identification of such features (concavo-convex regions along particle boundary), a new classification chart is proposed in this thesis to comprehend the kinematics of particles. The effects of critical parameters such as scale, resolution and user defined cutoff values on the quantification of shape parameters are analyzed and eliminated. The proposed methodology is compared with standard visual charts provided by earlier researchers and is demonstrated on real soil particles falling across a wide range of sizes and shapes. Finally, the role of particle shape in governing packing behaviour of aggregates is quantified based on the precise particle shape characterization.