大豆粗蛋白、粗脂肪含量近红外检测模型建立及可靠性分析  

Development and Reliability of Near Infrared Spectroscopy(NIS) Models of Protein and Oil Content in Soybean

作　　者：闫龙[1] 蒋春志[1] 于向鸿[2] 杨春燕[1] 张孟臣[1] 

机构地区：[1]河北省农林科学院粮油作物研究所,国家大豆改良中心石家庄分中心,河北省遗传育种重点实验室,河北石家庄050031 [2]中国农业科学院研究生院,北京100081

出　　处：《大豆科学》2008年第5期833-837,共5页Soybean Science

基　　金：国家重点基础研究发展计划(973计划)资助项目(2004CB117203);国家高技术研究发展计划(863计划)资助项目(2006AA10Z1B3、2006AA100104);国家支撑计划资助项目(2006BAD13B05)

摘　　要：大豆是重要的植物蛋白和食用油来源,目前常用的蛋白质和脂肪化学分析方法不仅操作步骤繁琐,而且籽粒粉碎而不能延续后代,在大豆品质改良的后代选育过程中的应用受到限制。为明确快速、简便、非破坏籽粒的近红外检测方法利用的可行性。利用4个建模样品集,采用偏最小二乘法建立了3个近红外大豆粗蛋白、粗脂肪籽粒检测模型和1个粉末检测模型。通过重新采集8个不同蛋白质和油份含量的大豆样品检测分析,并送检权威检测部门进行化学分析比较,分析近红外检测的可靠性。通过稳定性、一致性等分析表明,合适的建模样品集是正确建模的前提,在所建3个籽粒模型中,以含415个大豆材料的样品集所建模型(M6)可靠性最好;分析结果还表明,近红外检测结果与化学分析结果一致,在需要保存籽粒完好的大豆杂交分离世代或大量样品检测时,用近红外检测代替化学分析是可行的。Soybean provides a major source of protein and oil. The present analysis method of protein and oil is time- consuming and destructive, and its application in high oil and high protein breeding is restricted, hence, rapid and nondestructive analysis method is required for quality improvement. This research was done in order to validate the reliability of the near infrared spectroscopy（NIS） models which was rapid, simple and nondestructive. Near infrared spectroscopy of 4 sample pools were employed to develop regression models depending on least squares. They were 3 seed model and 1 powder model. Depending on re- sampling, comparison was done among 7 results including 3 chemistry results from different elite testing- center and 4 NIS models＇ results. The result showed that the suitable sample pool was necessary for excellent models. M6 which depended on a sample including 415 samples was the best one among 3 NIS seed model. Correlation among different methods were obviously consistent. Hence, using NIS models to substitute chemistry method was reliable especially for large scale nondestructive testing.

关 键 词：大豆 近红外 蛋白质 脂肪 

分 类 号：S565.1[农业科学—作物学]