Predicting sugarcane quality using a portable visible near infrared spectrometer and a benchtop near infrared spectrometer

Abstract

Sugar quality (Brix and Pol) is the key index to evaluate the value of sugarcane. Hence, a rapid, accurate, and time-efficient method is needed to determine the sugar quality. This study develops a two-point sugarcane quality model that uses a benchtop near infrared (NIR) spectrometer and a portable visible–near infrared (Vis-NIR) spectrometer to measure the sugarcane juice and stalk spectra, respectively. GT two experiments for developing a two-point sugarcane quality model. In the first, a model to calibrate the sugar quality as measured by a polarimeter and refractometer, and also by the benchtop NIR spectrometer. In the second, we developed a model to calibrate the sugar quality predicted from the calibration model developed in the first experiment, by measuring the sugarcane stalk absorption spectra using a portable Vis-NIR spectrometer. The results of the first experiment showed that the standard normal variate (SNV) spectral pretreatment was the most effective method for Brix calibration, with a coefficient of determination of prediction ( ${\mathrm{r}}_{\mathrm{p}}^2$ ) of 0.99 and root mean square error of prediction (RMSEP) of 0.2%. In the case of Pol, second derivatives were the best spectral pretreatment for effective calibration (r² = 0.99, RMSEP = 0.3%). The results of the second experiment showed that the multiple linear regression model developed using the stalk spectra with the second derivative was the best model for Brix calibration (r² = 0.70, RMSEP = 1.4%). The second derivative with the SNV pretreatment was best for Pol calibration (r² = 0.70, RMSEP = 1.4%). Our study showed that a sugar quality regression model can be developed for a portable Vis-NIR spectrometer using the data from the sugar quality predicted by a benchtop NIR spectrometer.