In Japan, mechanical sugarcane harvesting has been introduced in many regions. Although it improves theefficiency of sugarcane harvesting, it also brings the cane tops into sugar factories, which decreases yield and causeseconomic losses. In a previous study, a laser scanning technique was investigated for distinguishing cane tops from themechanically harvested raw sugarcane materials. This study improved the previous technique and developed a new patternanalysis method. A green He-Ne laser (wavelength 543.5 nm, output 4 mW) was used to scan the raw sugarcane materials.The back-reflected light intensity was measured by a light sensor (avalanche photodiode module). The surface conditions ofeach cane sample can vary in terms of the surface roughness and the substances that are adhered to it. Due to theirnon-uniform surface conditions, cane tops and cane stalks can be distinguished by analyzing the different distribution patternsof the back-reflected light intensity. During the experiment, 40 cane top and 40 cane stalk samples were used. The patternswere first analyzed by kurtosis and Fourier transformation, followed by linear discriminant analysis to classify these samples.The results showed that the overall accuracy for sample classifications reached 96.8%, of which the accuracies for cane stalksand cane tops were 95.7% and 97.8%, respectively.
Read full abstract