中国水稻科学

• 实验技术 • 上一篇    

用近红外反射光谱技术测定精米粉样品表观直链淀粉含量的研究

舒庆尧1;吴殿星1;夏英武1;高明尉1;Anna McClung2   

  1. 1 浙江大学 原子核农业科学研究所,浙江 杭州 310029; 2 Agricultural Research Service, Department of Agriculture, the United States of America, RT7, Box 999, Beaumont, TX 77713, USA
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:1999-07-10 发布日期:1999-07-10

Apparent Amylose Content of Rice by Near Infrared Reflectance Analysis of Ground Milled Samples

Shu Qingyao; Wu Dianxing; Xia Yingwu; Gao Mingwei; Anna McClung   

  • Received:1900-01-01 Revised:1900-01-01 Online:1999-07-10 Published:1999-07-10

摘要: 以约3 g精米粉为测试样品,研究了用近红外反射光谱(NIRS)技术测定稻米表观直链淀粉含量(AAC)的效果及影响校正的一些因素。对3种不同回归统计分析方法校正效果的差异比较表明,用修正的部分最小平方法(MPLS)、部分最小平方法(PLS)和主成分回归法(PCR)进行AAC校正时,校正标准误(SEC)、交叉检验标准误(SECV)和检验工作标准误[SEP(C)]均呈上升趋势,分别为0.83、1.75和1.09 (MPLS);1.73、1.98和1.74 (PLS)以及2.29、2.56和1.72 (PCR);相反,校正决定系数[WTBX]R[WTBZ]2和检验决定系数RSQ呈下降趋势,分别为0.983和0.91 (MLPS);0.927和0.84 (PLS)以及0.870和0.84 (PCR)。由此可见,用MPLS技术建立AAC回归方程效果最佳。而从两种校正群体选定方法对校正效果的结果看,用软件程序SELECT根据样品光谱特征选定的校正群体和用AAC测定值按比例选定的校正群体,在建立AAC回归方程时,前者的SEC、SECV和SEP(C)均比后者要小,而[WTBX]R[WTBZ]2和RSQ相仿,表明根据光谱特征选定校正群体效果略好。

关键词: 近红外反射光谱, 表观直链淀粉含量

Abstract: Using about 3 g ground milled rice testing samples, the effectiveness of apparent amylose content (AAC) analyzed by near infrared reflectance spectroscopy (NIRS) as well as the influencing factors for calibration were studied, whichwas indicated by the standard errors [[SEC for calibration, SECV for cross validation, and SEP(C) for validation ] and determination coefficients (R2 for calibration and RSQ for validation). Three different regression techniques, e. g. modified PartialLeast Square (MPLS), Partial Least Square (PLS), and Principal Component Regression (PCR), were compared for their effectiveness in calibration.The SEC, SECV and SEP(C) were o. 83, l. 75 and 1. o9 for MPLS; l. 73, l. 98 and l. 74 for PLS;and 2.29, 2. 56 and 1. 72 for PCR; While R2 and RSQ were o. 983 and o. 9l for MPLSI ;o. 927 and o. 84 for PLSl ;and o. 87 ando. 84 for PCR. These results indicated that MPLS is the best statistic method for calibration of AAC by NIRS. Two differentsample populations were compared for their efficiency of AAC calibration. one was selected by a computer program SELECTbased on NIRS characteristics, another was established by proportionally selection of samp1es with various AAC which wasdetermined by chemical analysis, these two populations were designated with P I and P Ⅱ, respectively. The SEC, SECV andSEP(C) for P I were smaller than that for PⅡ, while the R2 and RSQ were similar, which indicated that sample selection byspectroscopic character was more suitable for NIRS calibration.

Key words: near infrared reflectance spectroscopy, apparent amylose content, regression equation, rice