The protein content (PC) in rice is one of the indexes to evaluate the nutrition and taste quality of rice. Normal determination of the PC by chemical method is of high expense and time consuming. So the protein contents of samples with various dimensions were measured by the near infrared spectroscopy (NIRS) and the prediction models were established. It was discovered that the less the partical is, the stronger the capability of the prediction model. The correlation coefficient between prediction value of rice PC and chemical value is 0.94, the standard error of prediction (SEP) is 0.43, the mean relative error (MRE) is 2.1%, and the mean absolute bias is 0.21.%采用不同粒度的大米样品,用近红外光谱分析方法建立了其蛋白质含量的预测模型。大米蛋白质含量的模型预测值与化学测定值之间的相关系数达到0.94,预测标准差、平均相对误差及平均绝对偏差分别为0.43、2.1%和0.21。解决了传统的蛋白质化学测定方法检测时间较长的问题。
展开▼