Classification of normal and aged soybean seeds by discriminant analysis using principal component scores of near infrared spectra

Abstract

Near infrared transmittance spectra of normal and aged soybean seeds were measured using a specially designed sample cell for a single seed and were analysed by Principal Component Analysis (PCA) and Discriminant Analysis (DA) using PC scores for the classification of normal seeds and aged seeds. Soybean seeds were acceleratedly aged under an ambient condition of 42°C and 98% RH for 3, 5 or 7 days. As control some other seeds were also stored under condition of 15°C and 98% RH for 7 days, preserving the ability to germinate. After these treatments the seeds were dried at room temperature till they become initial water content. In multiplicative scatter corrected (MSC) spectra between normal and aged seeds somewhat changes were observed. The log (1/T) at 990 nm decreased with the increment of ageing days, while the log (1/T) at 1010 nm increased to the contrary. Aged seeds (7 days) were clearly separated from control using on the plane with the axes of second (PC 2) and third (PC3) principal components, derived from PCA of MSC spectra. The results of DA using 2PC scores (PC2 and PC3) which were effective to separate aged seeds, showed that 60% of aged seeds for 3 days, 80% of aged seeds for 5 days and 100% of aged seeds for 7 days were correctly classified. Phospholipids content in soybean seeds decreased with ageing, and then became half after 7 days, while total lipids content was constant. The loading plot for PC3 were similar to the spectrum of lipids, suggesting that PC3 reflected the changes of phospholipids during ageing. Consequently, it was concluded that NIR spectroscopy had a possibility for classifying normal and aged soybean seeds, and that the change of phospholipids content was an important factor for this classification.