The Research Of Tobacco Grading Based On Image Processing And Neural Network

Abstract:The quality of flue-cured tobacco and tobacco products has a direct relationship to the health of smokers. The existing methods for the classification of tobacco leaves of have major shortcoming as follow. The use of manual detection caused the subjectivity of grading is excessive, in addition, old methods lack of efficiency and accuracy. Attacking to these drawbacks, the computer image processing techniques is used for classifying the flue-cured tobacco according to their quality in this paper. In this paper, we choose tobacco leaf as the study object, and use computer image processing and neural network technology for automatic division of tobacco leaves. The major work as follows:(1)Tobacco image preprocessing method. On the basis of analysis of existing vector median filtering algorithm, a improved algorithm is proposed to carry out color smoothing filter processing of tobacco image. The major target is to delete noise that attaches in the image of tobacco. In addition, Iterative threshold method is utilized for tobacco image background segmentation, and contour extraction method is used for obtaining tobacco contour information.(2) Get the key parameters of tobacco. We choose the appropriate characteristics as a basis for judging the computer grading according to the classification standard of tobacco leaves. Get the tobacco leaf image’s color, shape and texture characteristics.(3)The realization of the classification algorithm. We analyze the relevance among main characteristic parameters of tobacco, a tobacco leafs classification method based on neural network is realized according to the true nature of its characteristic parameters.