Classification and Soluble Solids Content Prediction of Intact Tamarind Using Short-Wave Infrared Spectroscopy and...

The rapid, nondestructive assessment of tamarind quality is crucial for grading, pricing, and industrial standardization. Therefore, the purpose of this study is to classify and predict intact tamarind quality based on soluble solid content (SSC) by combining short-wave infrared (SWIR) spectroscopy with variable selection algorithms and chemometric models. SWIR absorbance spectra (900-1700 nm) were acquired from 120 samples of intact tamarind. Partial least squares-discriminant analysis (PLS-DA) and partial least squares-regression (PLS-R) models were developed using full spectra and various spectral preprocessing techniques to classify and predict SSC of tamarind, respectively, and the developed models were evaluated. The multiplicative scatter correction (MSC) and Savitzky-Golay (SG) second derivatives were selected as the best options for classification and prediction, respectively. Then, competitive adaptive reweighted sampling (CARS), the successive projections algorithm (SPA), and variable importance in projection (VIP) were applied to select effective wavelengths for developing simplified, more robust classification and prediction models for intact tamarind based on SSC. The PLS-DA-CARS model using MSC preprocessing achieved the best classification results with 97.2% accuracy, whereas PLS-R-CARS model with range normalization to predict SSC yielded a coefficient of prediction (R_p) of 0.85 and standard error of prediction (SEP) of 2.488%Brix.