[目的]利用近红外光谱技术构建定量预测模型,及时获得油茶重要的经济性状指标并进行评价,以简便快速、绿色安全而又准确地测定大批量油茶样品的含油率和脂肪酸组成,为油茶良种选育提供数据基础和依据.[方法]用气相色谱法测定 100 份油茶样品的含油率(Oil content,OC)及棕榈油酸(Palmitoleic acid,POA)、棕榈酸(Palmitic Acid,PA)、硬脂酸(Stearic acid,SA)、油酸(Oleic acid,OA)、亚油酸(Linoleic acid,LOA)、亚麻酸(α-Linolenic acid,ALA)、不饱和脂肪酸(Unsaturated fat acid,USFA)、饱和脂肪酸(Saturated fat acid,SFA)的含量水平,并作为试验测定值.对其中 90 份样本用近红外光谱技术,采用漫反射扫描方式,采集经过粉碎处理的油茶样品光谱,光谱序列通过光谱残差和特征残差结合的方法进行优化预处理,对光谱数据进行标准归一化,并结合导数法方法处理,结合偏最小二乘法建立油茶样品含油率和主要脂肪酸组成的定量预测模型.另外 10 份样本作为模型的外部验证.[结果]构建的油茶籽粉OC、PA、POA、SA、OA、LOA、ALA、USFA和SFA近红外光谱定量模型,建模集回归系数(RC)和验证集回归系数(RP)均在0.69以上,其中OC和LOA均达到0.90以上,模型效果较好.9个定量模型中效果最差的为SA,RC和RP分别为0.702 8 和0691 7.各指标定量分析模型中建模均方根误差(SEC)和验证均方根误差(SEP)的比值在0.80~1.26之间,其中ALA和USFA的SEC/SEP比值最接近1,分别为0.99和1.00,其定量模型较佳.油茶样品脂肪酸组成模型SEC/SEP比值的相对偏差在0.034~0.220,说明构建的定量模型结果很好.[结论]近红外光谱技术能实现快速、绿色安全和准确评价油茶样品的含油率和主要脂肪酸组成,用所构建的定量模型测定油茶籽粉的含油率和脂肪酸组成含量结果误差小,有较好的预测效果,在加快油茶等木本油料树种的良种选育方面具有很好的应用前景.
[Objective]In order to determine the oil content and fatty acid composition of large-scale Camellia oleifera seeds samples in a simple and rapid,green,safe and accurate method,a quantitative prediction model was constructed by near-infrared spectroscopy,and the important economic traits of C.oleifera were obtained and evaluated in a timely manner,which provided a data basis and basis for the selection and breeding of C.oleifera varieties.[Method]The content levels of oil content(OC),palmitoleic acid(POA),palmitic acid(PA),stearic acid(SA),oleic acid(OA),linoleic acid(LOA),linolenic acid(ALA),unsaturated fat acid(USFA),saturated fat acid(SFA)in 100 C.oleifera samples were determined by gas chromatography,and were used as test values.Near infrared spectroscopy(NIRS)and diffuse reflection scanning were used to collect the spectrum of 90 samples after composted treatment.The spectral sequences were optimized and pretreated by combining spectral residual and characteristic residual.The standard normalization of the spectral data was combined with derivative method and partial least square method to establish the quantitative prediction model of oil content and main fatty acid composition of the samples.Another 10 samples were used as external validation of the model.[Result]In the NIRS quantitative models of C.oleifera seed powder with OC,PA,POA,SA,OA,LOA,ALA,USFA,and SFA was constructed using partial least squares method,the regression coefficients of the modeling and validation sets were all above 0.69,with OC and LOA reaching above 0.90,indicating good model performance.Among the 9 quantitative models,SA had the worst performance,with the regression coefficients(RC)of the modeling set and the regression coefficients(RP)of the validation set being 0.702 8 and 0.691 7,respectively.The ratio of corrected root mean square error(SEC)and validated root mean square error(SEP)in the quantitative analysis model of each indicator was between 0.80 and 1.26,both closed to 1.Among them,the SEC/SEP ratio of C.oleifera seeds powder ALA and USFA was the closest to 1,with 0.99 and 1.00,respectively,indicating that the quantitative model was better.The relative deviation of the SEC/SEP ratio in the fatty acid composition model of C.oleifera samples was between 0.034 and 0.22,indicating that the quantitative model constructed had satisfactory results.[Conclusion]The NIRS can achieve rapid,green,safe,and accurate evaluation of the oil content and main fatty acid composition of C.oleifera samples.The quantitative model constructed has small errors in determining the oil content and fatty acid composition of C.oleifera seed powder,and has good predictive effects.It has good application prospects in accelerating the breeding of woody oil seeds.