Minireview on tobacco classification technologies: A vital bridge linking raw leaf properties with end product quality

Abstract

In the tobacco industry, classification serves as a vital bridge linking the properties of raw tobacco leaves with the quality of end products, playing a key role in ensuring product consistency and enhancing market competitiveness. This review systematically summarizes the latest progress from conventional analytical methods to cutting-edge intelligent systems in tobacco detection and classification. Four critical technological domains are systematically analyzed: (1) For appearance feature recognition, advanced image processing and deep learning technologies have enabled efficient, automated grading; (2) Regarding chemical structure detection, integrating various spectroscopic methods with machine learning has facilitated the precise identification of tobacco components; (3) Thermal reaction analysis coupled with machine learning, captures the distinct heat-release and mass-loss during pyrolysis and combustion, enabling accurate classification of different raw tobacco leaves; (4) The pre- and post-reaction product analysis methods have significantly enhanced both the accuracy and speed of complex component detection. Of particular note is the integration of in-situ hyperspectral detection with machine learning algorithms, which enables real-time, non-destructive classification and identification of tobacco leaves, while simultaneously providing detailed information on their chemical composition, such as nicotine and reducing sugars. The review concludes by forecasting future trends in tobacco classification technology, highlighting that the integration of intelligent, in-situ, and fully automated systems will enhance classification precision, shorten assessment cycles, and contribute to improved standardization in tobacco processing.