Determination of Main Constituents in Green Gram Using Near- Infrared Hyperspectral Imaging

Abstract

For the determination of main constituents, grain research laboratories around the world are using age old techniques which are time consuming, cost intensive, and sample destructive. In the present study, an attempt was made to investigate the feasibility of near-infrared (NIR) hyperspectral imaging for predicting moisture, protein, and starch content in green gram (Vigna radiata (L.) R. Wilczek). Images of green gram were obtained using a NIR hyperspectral imaging system in the wavelength region of 960-1700 nm at 10 nm intervals. Seventy five NIR reflectance intensities were extracted from each of the scanned images and were used in the development of prediction models. Ten-factor partial least squares regression (PLSR) and principal components regression (PCR) models were developed using a ten-fold cross validation for prediction. Prediction performances of PLSR and PCR models were assessed by calculating the estimated mean square errors of prediction (MSEP), standard error of cross-validation (SECV), and correlation coefficient (r). Overall, PLSR models demonstrated better prediction performances than the PCR models for predicting moisture, protein, and starch content of green gram. Based on β-coefficient values of the PLSR method, wavelengths regions of 1180-1220 and 1320-1360 nm; 960-980 and 1100-1110; and 1050-1100, 1230-1360, and 1400-1450 nm could be used in future inline inspection for predicting moisture content, protein, and starch content of green gram, respectively in multi-spectral imaging systems.