Near Infrared Spectrum Detection Technology Research In The Apple Juice Quality

Apple juice is the world’s second largest consumer goods, it has good natural flavor and high nutrition value. As a health drink, it deeply received the affection by consumers, and has a broad consumer market. Sugar, acidity, soluble solids are the main detection index of cider quality, but the existing detection methods exist the following problems:the operation is complicated, testing time is long and human factors is very big; Adulterated in the apple juice, apple juice concentrate of pesticide residues, both of them will affect the health of the consumer and restrict healthy development of juice industry.Therefore, it has become a hot spot to achieve rapid and nondestructive quality testing of apple juice in the apple processing industry and research field, and has great practical significance. Near infrared spectroscopy (NIR) analytical technique is a high efficient, rapid, non-destructive and green analysis technology which can be qualitative and quantitative. This paper adopts the near infrared spectral analysis technology combined with chemometrics for the detection of near infrared spectral analysis of sugar, acidity, soluble solids content in apple juice, to identify true and adulteration of apple juice, and detecting quantitative detection for the original juice content and pesticide arsenic in apple juice. It is of great importance to research to strengthen the supervision of the quality of apple juice and improve the corporate image, improve the level of apple juice quality detection.Research conclusions are as follows:1、Detection sugar, acidity, soluble solids content in apple juice by Near infrared spectrumThe best scattering treatment method for the standard of sugar is Standard Multi Scatter Correction. Mathematical treatment for derivative is1order, derivative processing interval points is4, smoothing interval points is4, not for a secondary smooth processing and partial least squares regression model is established for validation set RSQ is0.9490; The best scattering treatment method for the standard of acidity is Weight Multi Scatter Correction. Mathematical processing method for no processing, established partial least square method(MPLS) regression model to verify the set RSQ is0.9780; The best scattering of soluble solids on treatment methods is to deal with Standard Normal Variant plus to Detrend (S+D). Mathematical processing is for the second derivative every four points, smoothing interval points is4, not for a second smoothing and partial least squares regression model is established for validation set RSQ is0.9970.2、The identification, classification of the adulteration apple juice and juice content detectionThe scattering process is Standard Normal Variant plus Detrend for the classification and identification of3kinds of adulteration apple juice. Mathematical processing is1441, the best parameters c, g got respectively were11.3137,0.0039, the support vector machine (SVM) classification model established of training set correct classification rate is88.67%and the validation set correct recognition rate is83.34%.The best pretreatment method is no scattering processing for the content of testing three kinds of adulteration the original fruit juice and detection of adulteration HFCS raw juice content in apple juice samples. Mathematical processing is1441, the structure of BP neural network is established for the2-10-1after using principal component analysis method to extract2principal components. The model of training set correlation coefficient is0.974and the validation set the correlation coefficient is0.9349. The best pretreatment method for scattering processing is detrend only for detecting the original fruit juice content in adulterated FGS apple juice. Mathematical processing is1441, application of network performance function trainbfg back Propagation neural network is established after principal component dimension reduction. The model of training set correlation coefficient is0.90and the validation set the correlation coefficient is0.88. The best pretreatment method is Weight Multi Scatter Correction for detecting the contents of the original juice in adulterate sucrose apple juice. Mathematical processing is1441, the cross validation correlation coefficient (1-VR) for the principal component regression modelis established in0.9239and the validation set the correlation coefficient is0.990. Determined the pretreatment method is the Standard Multi Scatter Correction treatment combined with mathematical way0011for detecting the original juice content of three kinds of the adulterated sample. The structure of BP neural network is established for the2-10-1after using principal component analysis method to extract2principal components. The model of training set correlation coefficient is0.93271and the validation set the correlation coefficient is0.96742.3、The detection of Pesticide asomate residue in apple juiceScattering pretreatment method is NONE Only. Mathematical processing is1441, the structure of BP neural network is established for the5-10-1after using principal component analysis method to extract5principal components. The model of training set correlation coefficient is0.97543and the validation set the correlation coefficient is0.91809.