Non-Invasive Estimation Of Extraneous Matter Levels In Sugar Mill Inputs

Extraneous matter in sugar mill inputs consists of dirt and trash in billet cane. This study investigated the use of visual to very near infra-red (VIS/VNIR 400-1100nm) spectra of prepared cane to determine dirt levels and the use of image analysis to determine trash levels. VIS/VNIR spectra contain reflection/absorption features that are just as sensitive to dirt concentration as features in near-infra-red (NIR 1100-2500nm) spectra. VIS/VNIR spectra can be used to make similarly accurate dirt concentration estimates as NIR spectra. It is also possible to identify dirt type using VIS/VNIR or NIR spectra and knowledge of dirt type significantly improves the accuracy of dirt level estimates using either type of spectra. Image analysis of mixed billet cane and trash presented in this paper found that cane, leaf, and tops show characteristic patterns in the amount of light reflected in each band (R (red), G (green), B (blue), and NIR) but that there was considerable overlap between cane and trash categories. Including region based radiometric and spatial descriptors in cane/trash classification schemes improved classification accuracies to around 75%. Findings show that surface coverage proportions of leaf and cane for a given cane/leaf weight fraction are highly variable. Images of around one meter square are required to usefully discriminate trash levels for cane weight fractions between 80% and 100%.