In order to realize accurate and objective determination of the maturity grades of tobacco leaves. Middle part of tobacco leaves with different grades of maturity was used to extract color and vein characteristics with machine vision technique. Three color characteristics (H, S, V) and five vein characteristics (energy, correlation degree, entropy, contrast, inverse difference moment) were optimized by principal components analysis. Maturity grading models were built by back-propagation(BP)neural network. The result showed that the first four principal components together could represent the information of the three color characteristics and the five vein characteristics needed for grading, which realized the optimization of parameters. When the number of principal component factor was 4 and the number of nodes of hidden layer was 16, this grading model showed the best performance with average recognition rate of 93.67%. The overall results show that it is feasible to discriminate the maturity grades of fresh tobacco leaves with machine vision technique.