Utilization of near infrared spectroscopy coupled with chemometrics for non-destructive detection of mold in sweet tamarind

• Spectral measurement position influences classification model accuracy. • Average spectra from 6 measuring points provide the highest classification accuracy. • ANN model outperforms PLS-DA for classification accuracy and true positive value. Mold infection in sweet tamarind is a major postharvest internal fungal disease resulting in substantial output losses. Due to the fruit's lack of visible variations, mold-infected tamarind must be detected and classified non-destructively for postharvest management. This research demonstrated the efficacy of near-infrared spectroscopy in detecting mold infestation in tamarind, utilizing 400 samples with three designated positions that underwent interactance assessment of Fourier transform near-infrared (FT-NIR) between 800 and 2500 nm. Original spectral measurement of all positions impacts spectral intensity but not trends, as mold-infected pods exhibited higher spectral intensity due to enhanced absorption by moldy tissues. Four data preprocessing methods (standard normal variate; SNV, multiplicative scatter correction; (MSC), and first and second derivatives) were used to correct the spectral baseline and remove noises. Overall, the mean spectral model of three places, obtained from six integrated spectra, exhibited optimal performance, which was utilized to develop classification models employing principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), and artificial neural networks (ANN). PCA demonstrated to differentiate between molds, as both classes converged at their cluster center; however, overlaps persisted with ambiguous visual boundaries. The ANN model from SNV spectra outperformed the SNV-PLS-DA, providing the highest classification accuracy of 93.3% in the testing set. Furthermore, up to 97.6% of moldy pods were correctly classified. This study serves as a manual for using a streamlined, user-friendly, and economical method of full-wavelength NIR interactance spectroscopy for promptly identifying mold in sweet tamarind, enabling producers and food processors to minimize export loss and increase food security.