Using near infrared spectroscopy, a nondestructive mouldy chestnut detection method
Keywords:
NIR, supervised pattern recognition, nondestructive detection, chestnuts, mildewAbstract
One of the key agricultural crops, particularly in Asia, is chestnut. However, following harvest, mildew causes a significant loss of chestnuts. Finding and removing mouldy chestnuts is a crucial step in preventing and minimising widespread mildew. Additionally, this method is crucial in the manufacturing of food products made from chestnuts. In this study, we used near infrared spectroscopy to pinpoint mouldy chestnuts. A discriminating model was built using the near infrared spectra of 109 samples of chestnuts, including 40 samples without mildew, 40 samples with severe mildew, and 29 samples with mild mildew, spanning the wavelength range of 833 to 2500 nm.For the model's validation, a second set of samples with three mixed groups—chestnuts without mildew (n = 20), chestnuts with severe mildew (n = 20), and chestnuts with light mildew (n = 8)—was employed. The findings demonstrate that by utilising the first derivative and vector normalisation for spectra preprocessing and the Ward's algorithm as a distance algorithm approach, the best classification model based on the spectral band of 1818 - 2085 nm was attained. Sound chestnuts, somewhat mouldy chestnuts, and severely mouldy chestnuts were correctly classified at rates of 100, 92.8, and 100%, respectively. These findings showed that the near infrared spectral analysis-based discriminating model can successfully identify rotten chestnuts.