Relative performance evaluation of a custom-made NIR Reflectance Instrument and two commercial instruments (Foss and ASD) in the nondestructive moisture content measurement of in-shell peanuts

Abstract

A custom made Near Infrared Reflectance (NIR) spectroscope was used to determine the moisture content of in-shell peanuts of Virginia type peanuts. Peanuts were conditioned to different moisture levels between 6 and 26 % (wet basis) and samples from different moisture levels were separated into two groups marked as calibration and validation. NIR absorption spectral data, between 1000 nm to 2500 nm, was collected from the calibration and validation groups. Measurements were obtained on 30 replicates within each moisture level. Reference moisture data were developed using standard air-oven method. Partial Least Square (PLS) analysis was performed on the calibration set of spectral data and models were developed. The Standard Error of Calibration (SEC) and R2 of the calibration models were computed to select the best calibration model. Similar spectral data was obtained for the samples, from both the groups, using a Foss and an ASD spectroscope, and models were developed as done for the custom made instrument. The models were used to predict the moisture content of peanuts from the validation set, in each case. Moisture contents of the validation samples, as predicted by each instrument, were compared with their standard air-oven moisture values. Goodness of fit was determined based on the lowest Standard Error of Prediction (SEP) and highest R2 value obtained for the prediction models. The model, with reflectance plus normalization spectral data with an SEP of 1.57 and an R2 value of 0.96 compared well with that obtained with the ASD instrument (SEP =1.35, R2 = 0.88) and the Foss instrument (SEP =0.77, R2=0.97) for Virginia type in-shell peanuts.