Abstract: To explore the feasibility of predicting flue-cured tobacco yield based on RGB images in combination with machine learning and deep learning algorithms, a field experiment was carried out using flue-cured tobacco Zhongchuan 208. RGB images of flue-cured tobacco were obtained 25 days after topping using drones. Color, texture and shape of the images were extracted, totaling 35 features. Features selection were performed using the random forest algorithm, and a yield prediction model was constructed using seven machine learning algorithms (BPNN, GA-BPNN, ELM, PSO-ELM, SVR, GA-SVR, RF) and one deep learning algorithm (1D-CNN). The results showed as the follows: The accuracy (R²=0.970) and generalization ability (R²=0.817) of the random forest prediction model established by the combination of features (color, shape, and texture) obtained through the random forest algorithm are higher than those of the other six machine learning models and the convolutional neural network model. The predicted yield are in good accordance with the measured values. In summary, we constructed a tobacco yield prediction model through the combination of feature selection and the random forest algorithm and provided novel tools for tobacco yield prediction.