Research And Implementation Of Key Technology Of Tobacco Leaf Classification Based On Machine Vision

The quality of tobacco leaves directly determines the quality of cigarette tobacco. Tobacco leaves has been divided into 42 grades according to the national standards which based on tobacco maturity, leaf structure, color, length and other quality factors. Currently, there are plenty of shortcomings in the process of artificial tobacco acquisition such as a strong subjectivity and unified evaluation standard. The paper has researched the tobacco classification method based on machine vision inspection technology, developed a set of automatic recognition system of tobacco grade, and provided the basic support for realization of the tobacco leaf grading automation. This article main research contents are as follows:(1)This paper studied the influence of light source, camera, lens and other hardware devices on image quality, according to the physical characteristics of tobacco and the systems requirements, A image acquisition box of diffused positive illumination was slected.Finally a set of visble image acquisition system were built.(2)Due to signal interference noise of light and electromagnetic interference in the process of tabacco image acquisition, in order to get real high-quality images, image pre-processing of tobacco leaves was researched. 3×3 window median filter denoising and second order Laplasse operator was choosen to reduce the noise.(3)In order to extract correctly the characteristics of tobacco leaf image real-time processing and analysis, the color and length characteristic of tobacco leaves was used as the classification criterion.In the RGB and HSV color models, tobacco sample’s local area images were selected and then the leaf color characteristic parameters(R value, G value, B value, H value, S value, V value), were extracted by calculating the average value and variance, a library based on color feature level standard was established by the method of quantitative analysis of the color. Using external minimal rectangle extracts the leaf length feature value d parameters, a standard library was established based on the leaf length feature level.(4)According to the requirments of tobacco grading staff experience and vision detection technology, putting forward the fuzzy method in close to the largest degree the first time, in accordance with the first position(upper leaves, middle leaves and lower leaves)and then grading method. Sample standard of tobacco leaves grade feature vector template library was established by selecting tobacco features value of R, G, B, H, V and length d. Calculated close degree between the dected tobacco fuzzy vector and the sample standard grades of tobacco fuzzy set, using of the near-selection principle to obtain the maximum closeness, which identified tobacco grade.(5)Combined with image acquisition system, the author developed a tobacco identification system using VC ++2010 and Open CV2.4.4 development tools based on Windows software system, and has carried out a series of validation experiments to verify the rationality and feasibility of the algorithm.