Software for a particle-size analyser based on image analysis techniques

Abstract

This dissertation describes a technique of obtaining particle-size measurement using 2-dimensional image analysis. Images of coarse material on run-of- mine conveyor belts may be processed using the methodology described and size-calculation of particles is possible to a high degree of accuracy. The processing involves a histogram modification technique for performing image segmentation, a rule-driven procedure for merging regions and a region-splitting method to resolve particles that are joined together. Images of varying brightness and contrast may be processed without adjustment of any threshold levels, hence lending reliability and robustness to the approach. Particles may also be resolved from very indistinct background features formed by the finer material that is usually present. The work includes the software processing necessary to perform particle recognition on a given image frame providing a foundation on which a real-time instrument may be developed.