Automated Online Estimation of Fines in Ore on Conveyer Belt Using Image Analysis

Abstract

Excessive fines fed to metallurgical plants, power stations and other fixed or fluidized bed reactors can cause major problems by adversely affecting the flow of gas through the solid burden. In this industrial case study, the novel application of machine vision to monitor coal on conveyer feed systems is described. By using an adaptive thresholding and watershed transform, the fines in the stratified coal could be reliably estimated. We use the feature compactness to select and segment rock particles since crushed rocks tend to have round shape. The watershed transform is also used to split touching rock particles segmented from the adaptive thresholding.