Rapid detection of adulteration in desiccated coconut powder: vis-NIR spectroscopy and chemometric approach

Abstract

Adulteration of desiccated coconut powder (DCP) with coconut milk residue (CMR) is an emerging problem in the coconut processing industry. Consumers and industries are looking for a simple non-destructive device to measure the purity of DCP. vis-NIR (350–2500 nm) spectroscopy along with the chemometric techniques have been used to assess the purity of DCP. In this study, DCP was adulterated with CMR at different levels such as 0 (pure DCP), 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100% (pure CMR). Partial least squares regression (PLSR) models were developed using whole spectral data and selected wavelengths. The spectral data were pre-processed using different techniques such as raw, MSC + SNV, SG-smoothing, and detrending. The R 2 of the models constructed with the pre-processed spectral data was higher than 0.950, irrespective of pre-processing technique. Pre-processing of spectral data does not have a significant effect on model performance when compared with the model developed using raw spectral data (R 2 P = 0.973; SE P = 9.681; RPD P = 9.381; RER P = 10.389), but the prediction accuracy was decreased. The wavelengths 653, 933, 1189, 1383, 1444, 1670, and 1911 nm were selected as the featured wavelengths for quantification of adulteration level in DCP. No significant difference in statistical results was observed between the PLSR model developed with selected wavelengths (R 2 P = 0.869; SE P = 11.701; RPD P = 9.381; RER P = 8.595) and the PLSR model for whole spectral data. The developed model can be used to predict the level of adulteration in DCP if the adulterant concentration was more than 10%. The overall results obtained in present study suggest that the vis-NIR spectroscopy along with suitable chemometric techniques have a great potential for rapid measurement of adulteration level in DCP. • Vis-NIR (350–2500 nm) spectroscopy has been used to predict the purity of DCP • The wavelengths responsible for quantification of level of adulteration in DCP were selected. • The developed PLSR model can be used to predict the level of adulteration more than 10%. • The PLS-DA technique showed more than 70% classification accuracy.