A New Cluster Analysis-Marker-Controlled Watershed Method for Separating Particles of Granular Soils.

Md Alam,Asadul Haque

doi:10.3390/ma10101195

Abstract

An accurate determination of particle-level fabric of granular soils from tomography data requires a maximum correct separation of particles. The popular marker-controlled watershed separation method is widely used to separate particles. However, the watershed method alone is not capable of producing the maximum separation of particles when subjected to boundary stresses leading to crushing of particles. In this paper, a new separation method, named as Monash Particle Separation Method (MPSM), has been introduced. The new method automatically determines the optimal contrast coefficient based on cluster evaluation framework to produce the maximum accurate separation outcomes. Finally, the particles which could not be separated by the optimal contrast coefficient were separated by integrating cuboid markers generated from the clustering by Gaussian mixture models into the routine watershed method. The MPSM was validated on a uniformly graded sand volume subjected to one-dimensional compression loading up to 32 MPa. It was demonstrated that the MPSM is capable of producing the best possible separation of particles required for the fabric analysis.

Highlights

The separation of images of solid particles of granular soils subjected to boundary stresses is essential for studying the particle-level fabric
This paper presents a new cluster analysis-marker-controlled watershed method for separating solid particles of granular soils, named as the Monash Particle Separation Method (MPSM)
The algorithms developed in this study were implemented initially to assess less complex images prior to its application to relatively complex volumes of uniformly graded granular soils subjected to one-dimensional compression loading up to 32 MPa, which included stresses from pre- and post-crushing stress ranges

Summary

Introduction

The separation of images of solid particles of granular soils subjected to boundary stresses is essential for studying the particle-level fabric. The analysis of particle-level fabric requires separation of particles after segmentation. The importance of accurate separation of particles to study the fabric of granular soils has already drawn the attention of researchers [19]. The need for the maximum separation of particles for an accurate analysis of the particle-level fabric of granular soils has not yet gained an appropriate level of attention due to the inherent challenges envisaged in the development of separation algorithms. Inaccurate separation may result in erroneous fabric parameters, which may grossly undermine the benefits of particle-level fabric analysis

Methods

Results

Conclusion