Research Of Quantification Of Tobacco Leaf Grading Standard Based On Computer Vision

This paper is about quantification of tobacco leaf grading Standard under the supervise of computer vision technique, based on color theory, image segmentation, Gabor wavelet, grey theory and other technique, which achieved the measurement, analysis and description of leaf color, leaf shape, texture, maturity and oil quantitative. We do further research on feature extraction algorithms and setting parameters, and analysis the principles and methods of the leaf grading, that makes quantification of tobacco leaf grading standard close to reality.In this paper, we has designed the platform of tobacco leaf image gathering system based on the CCD, so the tobacco leaf image gathering was more convenient, accurate, and easy. Tobacco graduation will be divided into three comprehensive quantities integrated research such as color, leaf shape and texture. In this paper the color information of tobacco leaf was used a simple , easy, fast conversion HSV color space, making colors conversion from RGB to HSV can be completed in an instant. We extract 15 characteristic quantification parameters of HSV including the peak, energy, mean, variance and skewness etc. We used the "tracking insect" algorithm to track the border of tobacco image, and used step-tracking algorithm to track tobacco fitting image of the border. This method has a faster speed and consumes less time. It extracted the tobacco length, width, aspect ratio, area, perimeter, roundness, apex angle, disabling injury and other parameters. Using Gabor wavelet analysis, we has drawn the leaf tobacco texture average value and the standard variance; And we applied grey theory in leaf tobacco maturity and oil research firstly. It's through the parameters interrelated maturity and oil, based on the gray absolute interrelatedness, so obtained their varying degree quantification value.Finally integrating actual leaf tobacco graduation based on the GB2635-92.According to extracted parameters, we establish quantities model of tobacco leaf grading standard using first graduate then classification grading, so maked the tobacco more precisely.