Research And Implementation Of Tobacco Fuzzy Grouping Based On Computer Vision

According to the national tobacco quality classification standard,tobacco leaf can be divided into positive group and negative group.In the production process of tobacco products,negative tobacco is not meet the production requirements of tobacco leaves,it is necessary to separate the negative tobacco in the process.At present,the quality grouping of tobacco leaves is mainly depended on the artificial senses and experiences,which has strong subjectivity,low efficiency,and dose not adapt to the automatic production of tobacco.The method of tobacco automatic grouping based on computer vision has the characteristics of high speed and efficiency,which is the main research technique to solve the problem of low efficiency.It is recognized and valued in the industry.In addition,the idea of fuzzy clustering can solve the problem of unclear description of tobacco leaves.Therefore,it has great significance to apply fuzzy clustering method to tobacco leaf grouping.Aiming at the problem of strong subjectivity,low accuracy and efficiency,this dissertation uses the technology of computer vision to acquire and process the tobacco leaf image,and groups the tobacco leaf effectively by fuzzy clustering method.The specific contents are as follows:Firstly,the process of tobacco processing and feature extraction are improved.By comparing the different color models and combining the characteristics of tobacco leaves,we select the Lab color model,and set the threshold in the B channel for the tobacco leaf image segmentation.By using the wavelet transform algorithm processing to obtain wavelet coefficients of tobacco histogram,and using the gray level co-occurrence matrix to extract the inertia,energy,entropy and correlation of texture parameters,then we calculate the mean and standard deviation to provide the characteristic data for the grouping experiment.Secondly,a new idea of tobacco grouping is put forward.We divide the tobacco leaf into the front and the back tobacco leaves,then group these tobacco leaves respectively.Comparison analysis of the front and the back tobacco leaves and comparing the gray histogram,the characteristics of the front and the back tobacco leaves are different.Thirdly,a classifier based on fuzzy thought is designed.This dissertation selects the suitable membership function and fuzzy clear method,then deduces and analyzes the fuzzy matrix and clustering center.Finally,the front and the back tobacco of different batches were put into the classifier to carry on the fuzzy grouping experiment.Through the test,we can conclude that the tobacco fuzzy grouping method based on computer vision is effective,fuzzy rules can avoid the interference of human factors on tobacco grouping.The classifier based on fuzzy thought can automatically complete the experiment of tobacco leaf grouping with high accuracy.The front and the back information of tobacco leaf can be used as a new research method,which can further improve the correct rate of tobacco leaf grouping.