Fisher’s tobacco leaf grading method based on image multi-features

Abstract

To address the problems of manual tobacco grading, which is influenced by subjective factors and low accuracy of discrimination, this article discusses the automatic discrimination grading method based on machine vision technology. Firstly, a total of 16 image features were extracted from the geometric, color and texture classes of tobacco leaf based on the pre-processing of the collected tobacco leaf images. Next, the Fisher discriminant analysis model for tobacco leaf grade recognition was established with 16 image features from 38 groups of samples, and the accuracy of the Fisher discriminant analysis model was 97.4%. Finally, the other 7 sets of features were used as prediction samples to test the applicability of the discriminant model. The results show that the grading method has higher accuracy and stability compared with manual tobacco leaf grading, and can effectively identify the grades of small samples of tobacco leaf.