Study On Judgment Of Main Chemical Components Using Image Processing Techniques In The Curing Process Of Flue-cured Tobacco

In order to study the main chemical composition of flue-cured tobaccoin baking process nondestructive testing, based on the different stages in the process of flue-curedtobacco baking tobacco leaf as the research object, using image processing techniques to extractfresh tobacco leaf,38℃and42℃baking process,46℃,54℃and roast sample after thecolor of tobacco leaf characteristic value of hue (H) and saturation (S),(V) and brightness, texturecharacteristic value of angular second moment (W1), entropy (W2), deficit (W3), relevancy (W4),combined with the BP neural network prediction model is established, a preliminaryimplementation of different tobacco leaf baking stage of total sugar, reducing sugar, starch,nicotine and total nitrogen content of nondestructive testing. The main results were as follows:The color of tobacco leaf characteristic value brightness and hue (H)(V) are significantlycorrelated with the main chemical composition and saturation (S) was significantly correlated withtotal sugar and starch content, the hue (H) and the total sugar and starch was significantly negativecorrelated (P <0.01), and several other ingredients were extremely significant positive correlation(P <0.01); Brightness (V) and total sugar and starch content were extremely significant positivecorrelation (P <0.01), and reducing sugar, total nitrogen and nicotine showed a negativecorrelation, the correlation coefficient were-0.715,-0.786and-0.789.Tobacco leaf texture characteristic value of correlation with most of the chemicalcomposition is good, including angular second moment five chemical components weresignificantly correlated (P <0.01), entropy and starch were significantly correlated (P <0.05),with the rest of the four chemical components significantly (P <0.01), the gap spacing andcorrelation with the reducing sugar content was significantly correlated (P <0.05). Results showthat different baking stage main chemical components of flue-cured tobacco has close relationshipwith color and texture feature, to a certain extent, can be by color and texture characteristic valueof flue-cured tobacco to reflect changes in inner chemical components of flue-cured tobacco.Network predicted values and real values alignment is higher, the correlation coefficient of Rvalue in0.9above, the main chemical composition of the prediction accuracy is higher, and thereal value is very close. Maximum absolute error is0.97, the maximum relative error is5.81%;Total sugar of absolute prediction error is between0.08~0.97, the relative error between0.28~4.48; Reducing sugar of absolute prediction error is between0.02~0.38, the relative errorbetween0.14~1.90; Total nitrogen of absolute prediction error is between0.01~0.04, the relative error between0.79~1.79; Nicotine absolute error of the predicted values and themeasured values between0.01~0.05, the relative error is between0.47~2.13. Predicted resultsshow that the model according to the color of tobacco leaf characteristic value and texture featurevalues at different stages in the baking process of tobacco leaf total sugar, reducing sugar, starch,total nitrogen and nicotine five indicators forecast assessment is feasible, predicted results are ingood agreement with the actual situation, the model has certain practical value.