摘要
为实现红松仁脂肪无损、简便检测,利用近红外光谱分析技术对红松仁脂肪进行定量分析,用偏最小二乘法构建去壳红松仁脂肪定量分析模型,采用多种预处理方法优化模型,并且利用间隔偏最小二乘法、反向间隔偏最小二乘法、无信息变量消除法进行特征波段的筛选。结果表明,红松仁光谱经一阶导数预处理后建立的模型最佳;波段优选可以提升模型质量,其中反向间隔偏最小二乘法的筛选结果最佳,其松仁脂肪模型校正集相关系数为0.911 4,验证集相关系数为0.882 0,验证集均方根误差为0.646 8。可见,经过优化后,模型的预测性能较好,实现了去壳红松仁脂肪的快速、无损检测。
In order to explore a nondestructive and simple method to test the fat in peeled Korean pine seeds,near infrared spectroscopy was applied for the quantitative analysis of the fat. Partial least squares（ PLS） was used to establish the quantitative analysis models of the fat in peeled Korean pine seeds. Various pretreatment methods were used to optimize the models. Interval partial least squares（ i PLS）,backward interval partial least squares（ Bi PLS） and uninformative variables elimination（ UVE） were used to select characteristic bands. The results showed that,for the peeled Korean pine seeds,the model established after first derivative preprocessing had the optimal performance. The models could be promoted by the bands selection and Bi PLS was the optimization. The correlation coefficient of calibration subset of the fat models of peeled Korean pine seeds was 0. 911 4 and the correlation coefficient of predication subset was 0.882 0. The root-meansquare error of validation subset was 0.646 8. It was concluded that the model prediction performance was good and fast after optimizing,and nondestructive inspection of fat in Korean pine seeds was realized.
作者
仇逊超
QIU Xun-chao(Department of Computer Engineering,Harbin Finance University,Harbin 150030,China)
出处
《江苏农业学报》
CSCD
北大核心
2018年第3期692-698,共7页
Jiangsu Journal of Agricultural Sciences
基金
黑龙江省省属高等学校基本科研业务费基础研究项目(青年学术骨干研究项目)(2017-KYYWF-0089)
关键词
近红外光谱
红松仁
脂肪
定量检测
near infrared spectroscopy
peeled Korean pine seed
fat
quantitative detection
