[关键词]
[摘要]
本文采用拉曼光谱结合判别分析方法对新陈玉米进行了判别研究。在河南省内收集郑单958新陈玉米样品共75份,粉碎过筛后置于样品袋中,使用光纤直接采集他们的拉曼光谱。对原始光谱进行多项式平滑滤波、基线校正及一介导数处理后,首先运用主成分马氏距离判别分析方法建立了判别模型,主成分数为9,光谱建模范围为914~1369 cm-1时模型结果最优,此时建模集总正确判别率为92.7%,验证集总正确判别率为90%。然后运用偏最小二乘判别分析方法建立了相应的识别模型,当建模因子数为7,采用全谱建模时结果最优,此时建模集样品正确判别率为100%,验证集样品正确判别率为95%。偏最小二乘判别分析方法正确识别率较高,结果表明拉曼光谱可以应用于玉米新陈度的快速识别,在粮食储藏品质评价中具有极大的应用潜力。
[Key word]
[Abstract]
Fresh and stale corn samples were distinguished using Raman spectroscopy coupled with discriminant analysis. A total of 75 corn samples of the Zhengdan 958 variety were collected in Henan Province. After grinding and sieving, the powdered samples were placed in special PVC bags. Raman spectra were directly measured through the sample bags using optical fiber; polynomial smoothing, baseline correction, and first derivative methods were conducted to process the raw spectra. The discrimination model was developed with principal component discriminant analysis coupled with Mahalanobis distance. The best result was achieved when nine principal components were used with a spectral range of 914~1369 cm-1. The correct classification rates in the calibration set and the prediction set were 92.7% and 90%, respectively. Subsequently, the partial least squares discriminant analysis method was used to develop the corresponding recognition model. The best result was achieved when the number of factors was seven and the full spectral range was used. The correct classification rates in the calibration set and the prediction set were 100% and 95%, respectively. The correct classification rates obtained from the partial least squares discriminant analysis method were higher, indicating that Raman spectroscopy could be used to discriminate between fresh and stale corn rapidly and showed great potential in the quality evaluation of stored grains.
[中图分类号]
[基金项目]
河南工业大学高层次人才基金;郑州市科技局自然科学项目(20130925);863计划项目(2012AA101705)