Research On Online Classification Of Fresh Tea Leaves Based On Texture Analysis

With the rapid development of computer technology, the computer vision technology is widely used in processing,assortment and detection of food production. while few researches use the computer vision technology to classify and recognize fresh tea leaves. Nowadays, fresh tea leaves can be assorted roughly by traditional fresh-leaf classifier, but the classification accuracy is inaccuracy. As we know, the texture features of fresh tea leaves are distinct and easy to be extracted, so the classification of fresh tea leaves based on computer vision becomes a hot research area. The thesis’s contents and contributions are as follows:(1)Research of the preprocessing of fresh tea leaves. Firstly graying selects the weighted average method, and then dealing with the noise uses the median filtering.(2)Research of the method of texture feature extraction. According to the speciality of fresh tea leaves’s texture, the thesis adopts the Gray level co-occurrence matrix(GLCM) which is based on statistical method. Texture parameters include energy, correlation and contrast, homogeneity which are acquired through GLCM in different direction(00、450、900、1350) and different distance(10～60). We analyze the effect of these texture parameters.(3)Research of the classification of fresh tea leaves based on LS-SVM. This thesis adopts the method which is called Least Squares Support Vector Machine (LS-SVM) as the classifier of fresh tea leaves. The classifier adopts the inner product of Radial Basis Function (RBF) as the kernel function and uses violence test method to calculate the parameters of kernel function and the C parameter, the classifier also adopts adaptive Sequential Minimal Optimization (SMO) algorithm as training learning algorithm, and employes a one-to-one method for multi-class classification. Experiments show that the technology of GLCM and LS-SVM for classification of fresh tea can achieve very good results, the accuracy rate of classification can reach as high as96%. The researches of this thesis have high theoretical significance and practical significance in the online classification of fresh tea leaves.