【作者】 方彦；

【导师】 王汉宁；

【作者基本信息】 甘肃农业大学 ， 作物遗传育种， 2004， 硕士

【摘要】 玉米育种和品质筛选中要求分析技术准确而快速。现行的玉米营养成分检测，普遍采用实验室常规化学分析法，在试验工作中很不方便，而先进的检测仪器，不仅可以在短时间内分析样品，而且不破坏样品，被检测后的样品还可以用于种植，这将有利于加快玉米品质育种的进程。本研究所用近红外光谱仪正是随着光谱技术、计算机技术和化学计量学的发展应运而生的一种检测仪器，它具有测试速度快、操作简便等特点，与品质育种的要求十分吻合。但由于任何一台近红外光谱仪对每种组分或每种参数都要单独定标，因此，本研究初步探讨了HN1100型近红外光谱仪测定玉米品质的可行性。通过收集129份不同成分含量的玉米品种（系），将其分成两份，一份用于常规化学值的测定与粉末样品近红外光谱测定，另一份用于完整样品近红外光谱测定。在试验中先采用常规化学分析法测定玉米籽粒粗蛋白、粗脂肪和粗淀粉成分含量，并将其值作为真值，然后用近红外光谱仪在全光谱范围内每隔2nm采集一个数据点，共采集325个数据点，获得每个数据点处的吸光度值，在对不同光谱预处理与数学处理对建立定标方程的影响作了初步探讨后，根据定标集样品常规化学法测定结果和所扫描得到的光谱图间的拟合关系，采用偏最小二乘法估计有关参数，对完整籽粒样品和粉末样品分别建立定标模型，并用未参与定标的一组样品作为检验集对模型进行检验。结果表明：1.在全光谱范围内，玉米样品在不同的波长处有不同的吸收峰，这表明吸收强度与所测定成分的含量成正比。2.在所研究的玉米品质性状中，各性状近红外分析结果与常规化学测定结果之间有较好的相关关系，粗蛋白、粗脂肪和粗淀粉的检验集相关系数分别为0.937、0.945、0.964（粉末）和0.961、0.957、0.982（籽粒），并具有较低的标准误差（SEP），粗蛋白、粗脂肪和粗淀粉的SEP值分别为0.271、0.745、0.666（籽粒）和0.499、0.820和0.883（粉末），这表明利用近红外光谱技术测定玉米品质性状是可行的，能够用于育种的早代选择。3.针对玉米粉末样品，进行了光谱预处理对测定结果影响的讨论。供试样品经SNV、Mean Center、Basic Offset、Detrending、SNV+ Mean Center、SNV+ Basic Offset、SNV+ Detrending七种形式的光谱预处理后，各种参数与对照处理相比，均有不同程度的改善，这表明试验所用仪器的光谱预处理对定标结果和检验结果均有一定的影<WP=4>响。同时，在各种数学处理中，构建定标方程时应用不同的导数处理，其效果是不一样的。将各种光谱预处理形式相互组合建立定标方程，根据定标和预测后的各参数相比较选择最佳建模条件。其中，粗蛋白、粗脂肪和粗淀粉用于建模的最优光谱预处理条件为SNV+Detrending、SNV+Mean center、SNV+Mean center（籽粒）和SNV+Mean center、SNV+Basic offset、SNV+Mean center（粉末）；这几个指标的最优导数处理形式分别为：2阶、2阶、1阶（籽粒）与1阶、1阶、2阶（粉末）。更多还原

【Abstract】 Determining maize component is an important procedure in maize quality breeding. As usual, sample component was determined by normal chemical method that destroyed sample and very slow in some content. Advanced instrument can avoid the side effects.Near infrared spectroscopy is a kind of technique that can determine sample component quickly. But every instrument must calibrate singly. So this thesis is about the feasibility of an NIR system HN1100 for determining the quality of maize. 129 varieties which had different contents were collected and were divided into two groups, one used for determining chemical values and collecting NIR spectrum of powdered samples, the other used for collecting NIR spectrum of intact samples. The content of crude protein, crude fat, and crude starch was determined by normal chemical method, regarding the chemical values as true values. Reflectance reading were collected from 1100 to 1750 nm every 2nm. After approaching the effect of different spectra treatment and mathematical treatment to calibration, the method of partial least squares (PLS) analysis was used to model development for intact and powdered samples, respectively. The other samples were used to validation. The results showed as fallows: 1. The absorbing band of samples showed that maize had different absorbance in different wavelength region. There were direct correlation between absorbance and content of samples. So the spectra of maize can apply to quantitative analysis of maize.2. There were remarkable correlations between prediction values and chemical values of maize quality. The correlation coefficiencies were 0.937, 0.945, 0.964 (powdered samples) and 0.961, 0.957, 0.982 (intact samples) in validation set of crude protein, crude fat and crude starch respectively. The standard error validation (SEP) was 0.271, 0.745, 0.666 (intact samples) and 0.499, 0.820 and 0.883 (powdered samples) of crude protein, crude fat and crude starch respectively. This showed that it is feasible to measure quality of maize by NIR.3. Compared with control experiment, all the parameter were improved after scattering correction for spectra of samples by SNV, Mean Center, Basic Offset, Detrending, SNV+ Mean Center, SNV+ Basic Offset, SNV+ Detrending. This showed that scattering <WP=6>correction had significant effects on calibration and validation. Derivative treatments also affected the results. The optimal scatter correction condition of crude protein, crude fat, crude starch was SNV+ Detrending, SNV+Mean center, SNV+Mean center（intact samples）and SNV+Mean center, SNV+Basic offset, SNV+Mean center（powered samples）；The optimal derivative condition was 2、2、1（intact）and 1、1、2（powered）,respectively.更多还原