应用近红外光谱技术分析稻米蛋白质含量  

Analysis for Protein Content in Rice by Near Infrared Reflectance Spectroscopy(NIRS) Technique

作　　者：毕京翠[1] 张文伟[1] 肖应辉[1] 王海莲[1] 江玲[1] 刘玲珑[1] 万向元[1] 翟虎渠[2] 万建民[3] 

机构地区：[1]南京农业大学作物遗传与种质创新国家重点实验室,江苏南京210095 [2]中国农业科学院,北京100081 [3]中国农业科学院作物科学研究所,北京100081

出　　处：《作物学报》2006年第5期709-715,共7页Acta Agronomica Sinica

基　　金：国家高技术研究发展计划(863计划)(2003AA207020);国家重点基础研究发展规划(973计划)子课题(2004CB117200);国家"十五"科技攻关计划(2004BA525B02-04)

摘　　要：以稻谷、米粒、米粉3种形态的样品,应用近红外光谱技术(NIRS)和偏最小二阶乘法(PLS),建立了6个稻米蛋白质含量近红外光谱数学模型,并对模型预测结果的准确性进行了评价。结果表明,糙米蛋白质含量的稻谷、糙米粒和糙米粉近红外光谱预测模型校正决定系数(RC2)分别为0.893、0.971和0.987,校正标准差(RMSEC)分别为0.507、0.259和0.183;精米蛋白质含量的稻谷、精米粒和精米粉近红外光谱预测模型R2C分别为0.897、0.984和0.986,RMSEC分别为0.4970、.186和0.190。模型内部交叉验证分析表明,预测糙米蛋白含量的稻谷、糙米粒和糙米粉模型内部交叉验证决定系数(RCV2)分别为0.865、0.962和0.984,内部验证标准差(RMSECV)分别为0.557、0.290和0.205;预测精米蛋白含量的稻谷、精米粒和精米粉的模型RCV2分别为0.845、0.951和0.979,RMSECV分别为0.5940、.316和0.233。模型外部验证分析表明,预测糙米蛋白含量的稻谷、糙米粒和糙米粉近红外光谱模型外部验证决定系数(RV2)分别为0.683、0.801和0.939,外部验证标准差(RMSEV)为0.962、0.799和0.434;预测精米蛋白含量的稻谷、精米粒和精米粉近红外光谱的模型RV2分别为0.673、0.921和0.959,RMSEV为0.976、0.513和0.344。用米粉建立的近红外光谱预测模型准确性最高,米粒次之,基于稻谷的预测模型准确性相对较低;内部交叉验证和外部验证表明,近红外光谱分析技术与化学分析方法一致性较好,且能保证样品的完整性,在水稻优质育种和稻米品质分析中具有广泛的应用价值。Six predicted mathematic models for analysis of protein contents in brown and milled rice were established, with the technique of near infrared reflectance spectroscopy （NIRS） and partial least square （PLS） algorithm. The protein contents of brown rice and milled rice were determined by chemical methods. The different predicted models were established with the near infrared spectroscopy of paddy, whole grain and rice flour, which were to research the effect of sample forms on predicting veracity of NIR models. The determination coefficients （ Rc^2 ） of calibration of NIR models of brown rice protein content for paddy, whole grain and flour were 0.893, 0.971 and 0.987 respectively, the root mean square errors of calibration （RMSEC） were 0.507, 0.259 and 0.183. The Rc^2 of NIR models of milled rice protein content for paddy, whole grain and flour were 0.897, 0.984 and 0.986, the RMSEC were 0.497, 0.186 and 0.190 respectively. The veracity of models was estimated by the determination coefficients （ Rcv^2 ） and the root mean square errors （RMSECV） of cross-validation, the determination coefficients （Rv^2） and the root mean square errors （RMSEV） of external validation. The Rcv^2 of NIR models of brown rice protein content were 0.865, 0.962 and 0.984 respectively,the RMSECVO.557, 0.290 and 0.205. The Rcv^2 of NIR models of milled rice protein content were 0.845, 0.951 and 0.979,the RMSECV 0.594, 0.316 and 0.233. The Rv^2 of the NIR models for brown rice protein content were 0.683, 0.801 and 0.939,the RMSEV 0.962, 0.799 and 0.434. The Rv^2 of the NIR models for milled rice protein content were 0.673, 0. 921 and 0. 959, the RMSEV 0. 976, 0.513 and 0. 344. From the above results, we concluded that the veracity of flour NIR models is the best, that of whole grain better, while that of paddy poor. The NIR method can substitute the chemical method and be widely used in the breeding and quality analysis of rice.

关 键 词：水稻 蛋白含量 近红外光谱技术 校正模型 

分 类 号：S511[农业科学—作物学]