Improved evaluation of commercial cane sugar content in sugarcane stalk using near infrared hyperspectral imaging and stalk axis rotation technique

Determination of the commercial cane sugar (CCS) content in sugarcane stalk is an important process in the factory. A near infrared hyperspectral imaging system was investigated to develop a CCS prediction model directly from the stalk without peeling. Two scanning techniques and the presence of wax, or not, on the stalk surface were compared for prediction accuracy of the developed models based on laboratory analysis. The model was developed from the spectral data based on an axis rotation scan, which enabled rotation of the stalk around a stationary axis while being scanned. This model performed better than the model based on the data from a normal linear translation scan, especially for wax-covered samples. The axis rotation scan produced a pixel-wise CCS map that could be visualised across the whole surface of the stalk, whereas the pixel-wise CCS map based on the translation scan was able to show only the CCS values for a projected surface area and the opposite projected surface area of the stalk, with error at the curved edge of the stalk. In addition, the single colour CCS maps of each portion of the stalk, which were created by the model based on a translation scan of the wax-covered samples, could be used for visually detecting the right maturity of the sugarcane in the field.(c) 2022 IAgrE. Published by Elsevier Ltd. All rights reserved.

Automated summary

A near infrared hyperspectral imaging system was used to develop a prediction model for the commercial cane sugar (CCS) content in sugarcane stalk without peeling. The study compared two scanning techniques and the presence of wax on the stalk surface to evaluate the prediction accuracy. The model based on an axis rotation scan showed better performance, especially for wax-covered samples. The axis rotation scan generated a pixel-wise CCS map across the whole stalk surface, while the translation scan showed CCS values for a projected surface area of the stalk with errors at the curved edge. Additionally, the translation scan model using wax-covered samples could visually detect the maturity of sugarcane in the field.