Objective: A method based on RGB image processing to detect the blending proportion of cigarette stems and shreds was proposed to provide technical support for the optimization of cigarette blending uniformity. Methods: Tobacco powder was prepared by mixing leaf and stem into powder according to different proportion, and the RGB mean value of each test sample was determined by image analysis technology. The function model of stem blending proportion and RGB mean value was obtained by regression analysis of stem blending proportion and RGB mean value, and the accuracy, accuracy and repeatability of the model were verified. Results: The polynomial regression model of stem blending ratio and RGB mean value was established, and the fitting degree of the model was high, with the correlation coefficient of R² = 0.999 2; The relative error of the regression model was 0.27%～3.14%, the variation coefficient of the accuracy was 1.20%～2.02%, and the variation coefficient of the repeatability was 1.84%, which met the requirements of quantitative detection. Conclusion: A method based on RGB image processing was established to predict the blending proportion of cut tobacco stem. This method is simple and feasible, and more scientific and accurate than the traditional manual selection method.