Nondestructive Detection Method Based On Near Infrared Spectroscopy Buckwheat

With the rising of people’s living standards, people are no longer satisfied with the food adequate supply. People not only to. get diet nutrition in daily life, but also to achieve the purposes of health and illnesses. The buckwheat timely appear in people’s lives. At this stage, buckwheat quality primarily judge to subjective consciousness, it lack scientific quality detection and classification technology, so it is difficult to identify the Buckwheat quality and level quickly, it is result to poor quality of buckwheat and difficult to create high-quality buckwheat brand. The method of conventional buckwheat quality is rapid and efficient, but mostly due to personal subjective factors, so it is difficult to standardize. Therefore, we need fast, simple, accurate and simple operation way to change the backward situation of buckwheat quality testing in daily life. Near Infrared Spectroscopy (NIR) have many advantages, for example, rapid analysis, simultaneous determination of multi-component, without any pretreatment, non-destructive analysis, distance measurement, real-time analysis, the low cost of analysis and simple operation and so on. NDT method is fast alternative to conventional detection method in the agricultural products. Near Infrared Spectroscopy collected buckwheat spectral information to study the relationship between internal chemical properties and buckwheat quality. It establish fast, accurate, simple and stable way in the buckwheat quality evaluation, it has good theoretical significance and practical value. This thesis is carried out non-destructive testing method based on NIR technology on buckwheat. First, it establish buckwheat protein, starch and Total Flavonoids prediction model by near-infrared spectroscopy combine with PLS, Based on MATLAB platform using principal component analysis and BP neural network to predict buckwheat protein, starch and Total Flavonoid content; Then, using near-infrared spectroscopy and support vector machine algorithm to identify buckwheat species by buckwheat spectrum information; Finally, using mid-infrared spectroscopy combined with principal component analysis and BP neural network to establish buckwheat protein, starch and Total Flavonoids prediction model. In this paper, the main contents and conclusions are as follows:(1) FieldSpec3spectrometer collected buckwheat spectral information and to be pretreat with smooth and multiplicative scatter correction on buckwheat spectral information, it establish buckwheat protein, starch and Total Flavonoids linear and non-linear prediction model by near-infrared spectroscopy. First, it establish buckwheat protein, starch and Total Flavonoids prediction model by near-infrared spectroscopy combined with the PLS, but the buckwheat protein, starch and Total Flavonoids have poor correlation. And then using principal component analysis principal component extraction, the principal component scores with buckwheat protein, starch and Total Flavonoids content as the BP neural network input and output variables,it establish buckwheat protein, starch and Total Flavonoids prediction model. The conclusion is that buckwheat protein, starch and Total Flavonoids correlation value were0.718,0.761and0.911, the relative error is7.82%,4.35%and8.26%. The conclusions show that:it establish buckwheat protein, starch and Total Flavonoids prediction model by near-infrared spectroscopy combine with principal component analysis and BP neural network, due to Total Flavonoids has high correlation, so it can achieve the purpose of buckwheat Total Flavonoids predicted, using PLS combined with near-infrared spectroscopy to predict buckwheat protein, starch and Total Flavonoids contention is not feasible..(2) Using near-infrared spectroscopy and support vector machine algorithm to identify buckwheat species with buckwheat spectral information, the LDBSVM package can identify the varieties of buckwheat on MATLAB platform, it trained seven support vector machine and the eight different species of Origin buckwheat predictive correlation value of92.5%on average. The conclusions show that:using near-infrared spectroscopy and support vector machine algorithm to identify the spectrum specie s of buckwheat is feasible.(3) Based on mid-infrared spectroscopy combined with principal component analysis and BP neural network to establish buckwheat protein, starch and Total Flavonoids prediction model, The conclusions show that:buckwheat protein, starch and Total Flavonoids forecast correlation value are0.769,0.848and0.938, the relative error is9.31%,4.53%and8.53%, it due to starch and Total Flavonoids have high correlation, so it can achieve the purpose of buckwheat starch and Total Flavonoids predicted. so it can achieve the purpose of buckwheat starch and Total Flavonoids predicted. Relative to the near-infrared spectroscopy, mid-infrared spectroscopy prediction accuracy is higher than the prediction accuracy of near-infrared spectroscopy, infrared spectroscopy is a lossy detection is not conducive to the rapid detection of agricultural products.In summary, this paper discusses the component prediction and species identification based on near-infrared spectroscopy combined with stoichiometry buckwheat. In the buckwheat component prediction, using near-infrared spectroscopy combined with PLS and BP neural network to predict buckwheat protein, starch and Total Flavonoids content. In the buckwheat identification, using near-infrared spectroscopy and support vector machine algorithm to identify buckwheat spectral information. Based on mid-infrared spectroscopy combined with principal component analysis and BP neural network to establish buckwheat protein, starch and Total Flavonoids prediction model. It designed to improve the of buckwheat near-infrared spectral detection method, The results of this study with independent intellectual property rights in the buckwheat and intrinsic quality of detection methods in the other agricultural products, it provide basic theory and method and has a good theoretical significance and practical value.