Critical quality attributes determination of flue-cured tobacco based on NIR combined with variable selection approach optimization

Abstract

The critical quality attributes (CQAs) of flue-cured tobacco are closely related to overall quality. It is crucial to develop rapid and precise methods for determining the total sugar (TS), reducing sugar (RS), and total nitrogen (TN) of flue-cured tobacco. In this study, we established and optimized quantification models using near infrared (NIR) spectroscopy combined with variable selection methods, including a proposed single correlation regression parameters algorithm (SCRPA), to extract informative features from single correlation regression parameters. The overall results demonstrate that SCRPA is a promising method with improved prediction performance (RMSEP values of 0.4399% for TS, 0.8733% for RS, 0.0233% for TN) and smaller bias (0.2234% for TS, 0.9483% for RS, 0.0227% for TN). Moreover, NIR spectroscopy combined with variable selection methods could be efficiently employed to monitor the quality of flue-cured tobacco and is suitable for on-site analysis, facilitating the rational use of tobacco resources.