Comparative nondestructive measurement of corn seed viability using Fourier transform near-infrared (FT-NIR) and Raman spectroscopy

The commercialization of agriculture has driven the need to ascertain the quality of agricultural inputs, especially seeds in order to optimize output and increase economic returns. Seed viability is a critical consideration for ensuring a reasonably high harvest. More often than not, farmers experience losses after a significant percentage of seeds fail to germinate after planting. The loss of seed viability may be due to a number of reasons such as overheating during drying, physical damage during post-harvest processing, and ageing during storage. It is therefore critical for seed companies to sufficiently inspect their products and uphold them to acceptable seed quality standards in order to gain credibility and ensure business sustainability. In this study, the Fourier transform near-infrared (FT-NIR) and Raman spectroscopy techniques were used for evaluating seed viability to investigate their comparative advantages with regard to the corn viability test and classification. The techniques were applied to white, yellow, and purple corns with 300 samples in each category. The 300 sample corn seeds were divided into two groups of 150 seeds each; one group was heat-treated using microwaving, and the other was used as the control. Sample spectra from treated and untreated corn seeds were collected using an FT-NIR spectrometer in the wave range of 1000–2500 nm, and then Raman spectrometer in the wave range of 170–3200 cm−1. The collected spectra were divided into training and testing sets, corresponding to 70% and 30% of the total, respectively for calibration and validation of the techniques. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were used to assess the spectral data from FT-NIR and Raman spectroscopy. The analysis results indicated that FT-NIR spectroscopy correctly classified viable and nonviable seeds for all the three categories of corns with a high accuracy of 100% and a predictive ability of more than 95%. Moreover, Raman spectroscopy demonstrated reasonably high classification accuracy with PLS-DA, but a significant number of seeds were overlapping when using PCA. In addition, an analysis of variance (ANOVA) indicated that the difference between treated and untreated corn seeds was not statistically significant (P < 0.05). The study demonstrated that FT-NIR spectroscopy is superior to Raman spectroscopy in evaluating corn seed viability.