Robust NIRS Models for Non-Destructive Prediction of Physicochemical Properties and ageing of Basmati Rice

Abstract

Aim: To determine physicochemical properties and age of rice by non-destructive technique. Place and Duration of Study: Study was conducted at Division of Food Science and Postharvest Technology, Indian Agricultural Research Institute, New Delhi during 2020 to 2021.  Methodology: Rice were kept for accelerated aging at 42.6°C temperature & 71% RH for a duration of 30 days. Changes in four physicochemical properties namely amylose content, volume expansion ratio (VER), water absorption ratio (WAR), and kernel elongation ratio (KER) were evaluated destructively (by spectrophotometer and cooking method) and non-destructively (by spectroradiometer) at every alternate day, during 30 days storage. Results: The physicochemical parameters of rice showed a good correlation with spectral signatures.  Subsequently, Principal component Analysis (PCA), Partial Least Square Regression (PLSR), and Multiple Linear Regression (MLR) were used to model the physicochemical changes occurring during the process of accelerated aging using spectral reflectance values. Based on values of Coefficient of determination (R²) and Root mean square error (RMSE) accuracy of models was determined. Predictions with the MLR model resulted in a coefficient of determination (R2) of 0.82, 0.87, 0.9,7, 0.83 and 0.82 with root mean square error (RMSE) of 0.18, 0.13, 0.21, 0.124 and 4.2 for amylose content, VER, WAR, KER, and ageing process respectively for calibration. Conclusion: The study demonstrated the potential of NIRS in non-destructively predicting the physiochemical parameters of rice.