Near Infrared Spectral Quantitative Model of the Nutrient Content in Rice
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
Protein, fat, carbohydrate and moisture in rice are important nutrients in rice. The detection of these components needs to rely on the national standard method. The detection process is tedious and has large pollution. In this experiment, 75 rice samples were studied, and their contents of protein, fat, carbohydrate and moisture were measured by national standard method. Spectral information was collected by AntaresⅡ infrared spectrometer. Samples were divided into verification set and correction set, including 63 samples in correction set and 12 samples in verification set. A quantitative model of protein, fat, carbohydrate and moisture of rice was established through six spectral pretreatment methods and three modeling methods, and the model by running workflow was used to realize rapid detection of four components. The results showed that proteins and carbohydrates followed by the first derivative and Savitzky-Golay filtering, while fat and water followed by the second derivative and Savitzky-Golay filtering. The quantitative model of protein, fat, carbohydrate and water content in rice was established by Least partial square method in combination with the measured values by national standard method and has high prediction accuracy. The internal cross-validation coefficient of the prediction model of protein, fat, carbohydrate and moisture is 0.9266, 0.9333, 0.9198, 0.9175, and RMSECV is 0.280, 0.133, 0.362 and 0.270, respectively. The internal cross-validation R value is above 0.91 and RMSECV is less than 0.35, which can realize the rapid detection of protein, fat, carbohydrate and moisture in rice.
