Application Of Image Processing Technology And Support Vector Machine In Tobacco Grading

Tobacco is an important economic crop. Its quality is the basis of cigarette products. In many countries and regions, the tobacco industry has a pivotal position in the whole national economy. If we want to improve the quality of tobacco products, we should have scientific and rational leaf quality standards and accurate grading identification.Tobacco classification is based on experience in our country. There are no specific quantitative targets. The quality is still determined by the sense of taste and smell of professional who has received special training. With the development of intelligent classification technology, now more than by using artificial neural network, pattern recognition, etc for automatic grading of tobacco leaf samples images, but the accuracy of classification is not very satisfactory. For the low efficiency and accuracy in the tobacco manual detection and grading, the computer image processing technology and the method of support vector machine (SVM) is introduced into the flue-cured tobacco leaves in the automatic classification.This article mainly has carried on the following three aspects:1. Discussions on the digital image processing technology, and make full use of digital image processing technology in tobacco leaf image processing. According to the characteristics of the acquisition of tobacco samples, choosing a suitable image segmentation method for tobacco leaf. The improved vector median filter method is chose to smooth the color tobacco image and avoid the noise in it. The mathematical morphology and edge detection method of combining the LOG operator is put forward to improve the effect of edge detection. It has laid a solid foundation for extracting the feature parameters of the tobacco.2. Saturation and brightness mean value are extracted as the color feature parameters of tobacco using the HIS color model which is close to human vision. The Least-square elliptic fitting method is applied to extract the length, width, ratio of length and width, area, perimeter, etc of tobacco leaves as the shape characteristic parameters. Because of the tobacco leaf texture characterization of the indices such as the oil content of tobacco leaves, using the methods of Gabor texture analysis to extract the leaf texture parameters.3. Considering the characteristic parameters of nonlinear tobacco leaves and the shortage of the existing classification method, Support vector machine (SVM) algorithm was proposed to build model of levels of tobacco leaf, tobacco leaf type recognition based on the SVM technology is effective and feasible by simulation experiment. It shows that the superiority of the SVM used for tobacco classification field according to by compared with the BP neural network tobacco classification model.It shows that the method of leaf grades recognition rate has greatly improved, Compared to determination of chemical composition or spectral classification, this method is a kind of nondestructive testing, and the cost is low, in the actual purchase of tobacco leaf has stronger practicability, can obviously increase the speed, the classification result is more accurate and objective.