Aiming at grading flue-cured tobacco leaves, an intelligent method based on rough set theory was proposed. Traditional rough set theory was generalized for the characteristics of flue-cured tobacco leaves’grading, and relevant algorithms of discretization and attribute reduction based on rough set theory were also given. In order to obtain knowledge with reasonable granularity, a two-level reasoning model that consists of grouping and grading was constructed. By parallel grouping according to leaves’positions and colors based on chemical compounds, inference rules and importance degree of each chemical compound can be obtained. Then, through multi-attribute evaluation found on the importance degree and the group inferred from the rules, final grade of the given flue-cured tobacco leaf was determined. Experiment results prove that this method is effective and credible, and has advantages over the previously proposed methods.