Use of near infrared spectroscopy for rapid and non-destructive identification of morphologically similar seeds of different paddy varieties [Oryza sativa L.]

Rice production systems and breeding programmes often require handling large numbers of paddy samples, where maintaining sample identity is critical. Conventional identification methods, such as morphological and molecular analyses, are reliable but laborious and time-consuming. To address this, the present study evaluated the use of near infrared (NIR) spectroscopy (588–1091 nm) combined with soft independent modeling of class analogy (SIMCA) for rapid discrimination of paddy seed varieties. A total of 750 seed samples representing five varieties (Bg300, Bg352, Bg357, Bg359, and Bw363) were scanned, and SIMCA prediction models were developed using Pirouette 3.11 software. Model parameters were optimized to enhance classification performance. The influence of moisture content (MC) at four levels (10%, 12%, 15%, and 23%) during seed handling was also investigated. The SIMCA models achieved variety-specific classification accuracies of 89% (Bg300), 83% (Bg352), 88% (Bg357), 87% (Bg359), and 84% (Bw363). When averaged across all varieties, correct identification rates were 82% at 10% MC, 83% at 12% MC, 88% at 15% MC, and 92% at 23% MC. The results confirm an increasing trend in predictive accuracy with elevated MC, attributable to stronger variety-specific water-related absorbance features in the NIR region. Overall, the study demonstrates that NIR spectroscopy combined with SIMCA provides a rapid, non-destructive, and feasible approach for varietal identification of paddy seeds, particularly effective under equalized moisture levels. This method holds strong potential for field-level application, offering faster and more reliable variety authentication compared to conventional practices.