Study On The Method Of Computer Vision Technology Using For Controlling Fermentation Of Black Tea

Based on reviewing the applications of computer vision technology in agricultural, this paper researched on the methods of using computer vision technology to controlling fermentation of black tea, and discussed the problems of the method in the experiment applicated in production.1) Computer vision hardware system for researching of black tea fermentation is established, including the selection of camera, light source, and designing of the black tea fermentation tank and photo frame. The hardware system is relatively stable, the stability of the sensitivity of color and light intensity are preferably.2) This study choosed the MATLAB image processing module, the color appearing in the picture was roughly divided into three part:first category color (green), the second color (red), the third type of color (blue and black color), using the experience formula of the RGB values to determine the classification of three kinds of color:The classification effect is preferably. Using RGB values to determine the characteristics of three kinds of color segmentation method, gray level is less than 75 pixel area for the third type of color; Artwork after separation of the third kind color, hue value less than or equal to 0.67 area for the first kind of color; Greater than or equal to 0.67 to the second color.3) Changes of six groups of characteristic value of red broken tea and congfu black tea fermentation processing are described, are OE (Organoleptic Evaluation),TP (Tea Polyphenol),TF (Theaflavins) L*、a*、b* (L*、a*、b* Value of Tea Liquid), and analysised the correlations with 21 characteristic values(R, G, B, H, S, I value and Z-the size of each Three colors’ area) extracted from two kinds of black tea images, the tea quality and physical and chemical screening of the correlation between larger eigenvalues linear regression model is established. The fitting degree reached 80% significance level is 0.05 and model, the time black tea and red tea three sets of each item. Time income model for black tea,Time income model for black tea, OE=-91.367+11.823H2,R2=0.969; TP=16.672-21.940S3, R2=0.836;L*=-48.201+71.115S3+157.826I3, R2=0.864 Red broken tea income model for TF=-4.642+27.193I3,R2=0.837; OE=45.558-115.517S1, R2=0.894; L*=71.021-103.248S3, R2=0.820;a*=29.752-0.00003Z2,R2=0.800.4) Using each 12 groups of red broken tea and congfu black tea image eigenvalue, respectively established black tea fermentation degree of discriminant model which is based on BP neural network, the results showed that the discriminant accuracy of two models are respectively 81.25% and 87.5%.