Non-destructive Prediction of Brix Value in Sugarcane Based of Portable NIR Spectroscopy

Abstract

The quality of sugarcane in the plantation is the most important information for farmers and sugar factories for assessing the maturity of sugarcane and determining the optimal harvest schedule. The brix value is used as a quality index in the sugar industry and is an important parameter for the evaluation of cane quality and maturity. Traditional methods of determining brix involve time-consuming and labor-intensive processes, often involving destructive sampling. To overcome these challenges, this study proposes a non-destructive approach using portable near-infrared (NIR) spectroscopy to predict the sugar content in sugarcane stalks. The main objective of this study was to develop a nondestructive prediction model for the brix value in sugarcane using portable NIR spectroscopy. Data processing involved two models: Partial Least Squares (PLS) and an Artificial Neural Network (ANN), along with various data pre-treatment techniques. The PLS model showed an improvement in prediction accuracy with data pre-treatment, especially with the Savitzky-Golay method (R² = 0.755, RMSEP = 1.22%, RMSEP = 1.43%, CV = 6.13%, and RPD = 2.02). In addition, the ANN model combined with Principal Component Analysis (PCA) showed high predictive performance when sugarcane was 11 months old (R² = 0.797, RMSEC = 0.56%, RMSEP = 0.87%, CV = 3.04%, and RPD = 2.96).