高光谱成像技术在烟叶和杂物分类中的应用

Application of hyperspectral imaging technology to tobacco leaf and foreign matter sorting

为提高卷烟产品品质,避免烟叶中掺杂异物,在高光谱成像技术的基础上,结合Savitzky-Golay平滑滤波(SG)、多元散射校正(MSC)、支持向量机(SVM)建立了烟叶和杂物的分类方法。样品分类精度采用总体分类精度(OA)和卡帕系数(Kappa)进行表征。实验结果表明,使用径向基函数对样品进行分类时效果最佳,烟叶和杂物总体分类精度为99.92%,卡帕系数等于0.998。基于高光谱成像技术的烟叶杂物分类方法可以准确分辨烟叶、塑料橡胶制品和金属制品等。

In order to improve the quality of cigarettes and to reject foreign matters from tobacco leaves during cigarette manufacturing,a tobacco leaf and foreign matter sorting method was established on the basis of hyperspectral imaging technology combined with Savitzky-Golay(SG)smoothing filter,multiple scattering correction(MSC)and support vector machine(SVM).The sorting accuracy was characterized by the overall sorting accuracy(OA)and Kappa coefficient.The experimental results showed that the sorting effect was the best when radial basis function was adopted.The overall tobacco and foreign matter sorting accuracy reached 99.92%with the Kappa coefficient of 0.998.The sorting method based on hyperspectral imaging technology can accurately distinguish plastic,rubber or metal products from tobacco leaves.