用近红外光谱技术快速测定籼稻品种的蛋白质含量  

Rapid Determination of Protein Content in Long Grain Rice by Near Infrared Spectroscopy

作　　者：俞法明[1] 陆艳婷[1] 严文潮[1] 刘庆龙[1] 金庆生[1] 

机构地区：[1]浙江省农业科学院作物与核技术利用研究所,杭州310021

出　　处：《中国粮油学报》2009年第5期134-138,共5页Journal of the Chinese Cereals and Oils Association

基　　金：浙江省科技攻关项目(011102471)

摘　　要：分别以稻谷、糙米、精米和精米粉为扫描材料,应用近红外光谱法建立了籼稻蛋白质含量的预测模型。结果表明,采用光谱预处理的校正效果比不采用预处理的好,用偏最小二乘法(PLS)获得的籼稻稻谷、糙米、精米、精米粉的回归模型和交叉验证结果为:最优校正决定系数(R2)和交叉检验均方误差(RMSECV)分别为0.772 10、.507,0.888 40、.379,0.911 6、0.336,0.951 0、0.258,稻谷的误差最大,粉样的误差最小。育种实践中,低世代可选用糙米、高世代可选用精米和精米粉作为扫描样本测定稻米蛋白质含量。Prediction models for protein contents of long grain rice were established with near infrared spectroscopy using paddy rice, brown rice, milled rice and milled rice flour as scanning materials. Results show that pre - spectral treatment gives better calibration effect than directly treatment. The regression models derived from the partial least squares （PLS） for paddy rice, brown rice, milled rice and milled rice flour, as well as the cross - certification result as follows： the optimal calibration determination eoefficient （R2） and the crossexamination mean square errors （RMSECV） for paddy, brown rice, milled rice, and mired rice flour are 0.772 1, 0.507; 0.888 4, 0.379; 0.911 6, 0.336; and 0.951 0, 0. 258, respectively. Absolute error is less by using milled samples, compared with using paddy. In terms of the prediction models, it is proposed that brown rice could be used as scanning sample when earlier generation options are done, and milled rice and milled rice flour could be used while higher generation options are done in rice breeding practice.

关 键 词：近红外光谱法 籼稻 蛋白质含量 快速测定 

分 类 号：S511.21[农业科学—作物学]