Optimization of fresh flue-cured tobacco maturity discrimination model based on machine vision

In order to solve the problem of inaccurate judgment of fresh tobacco maturity based on subjective experience, image processing and feature extraction were carried out on fresh tobacco leaf with different maturity, and a fresh tobacco leaf maturity judgment model was established to realize intelligent judgment of fresh tobacco maturity. By collecting 10 color features and texture features of upper leaf images of Yunyan 87 varieties with different maturity, variable cluster analysis and correlation analysis were carried out respectively, and the one feature with the strongest correlation between each type of feature and maturity was selected to form a feature subset. In this paper, support vector machine based on genetic algorithm（GA-SVM）, Back propagation neural network based on particle swarm algorithm（PSO-BP） and Extreme learning Machine（ELM） were used to identify the maturity of fresh tobacco. The results showed that the optimal five tobacco leaf image features were used as model inputs, the discrimination accuracy of the established GA-SVM、PSO-BP、ELM model was 92.00%、90.00%、84.00%. It is proved that it is feasible to discriminate the maturity of fresh tobacco by using machine vision technology, which provides theoretical basis and technical support for intelligent tobacco leaf harvesting.