|
Components Discrimination for Formula Tobacco Based on Hyperspectral Imaging |
|
View Full Text Download reader |
DOI: |
KeyWord:formula tobacco components discrimination near-infrared hyperspectral imaging pixel-wise classification principal component analysis successive projection algorithm second deri? vate method |
Author | Institution |
MEI Ji-fan, LI Zhi-hui, LI Jia-kang, SU Zi-qi, XU Bo,DU Jin-song,XU Da-yong,LI Hua-jie |
1. Key Laboratory of Tobacco Processing Technology,Zhengzhou Tobacco Research Institute of CNTC,Zhengzhou ,China;2. China Tobacco Fujian Industrial Co. ,Ltd. ,Xiamen ,China |
|
Hits: 1297 |
Download times: 1069 |
Abstract: |
Near-infrared(1 000-2 200 nm)hyperspectral imaging technique was applied to the dis? crimination of components in formula tobacco,including cut lamina,cut stem,expanded tobacco and reconstituted tobacco. Two approaches,named pixel-wise and object-wise,were investigated to con? duct this research. The pixel-wise components discrimination study was based on the spectral data of all pixels of the hyperspectral images of samples. Second derivative coupled with Savitzky-Golay (SG)algorithm was applied as preprocessing method for original spectral data. Through principal component analysis of the pixel data,the feasibility for component discrimination of the pixel hyper? spectral data was confirmed. The established support vector machine(SVM)model based on first five components' data showed its excellent character in discriminating cut lamina and cut stem,cut lamina and reconstituted tobacco,obtaining intuitive discrimination results. The K-nearest neighbor and sup? port vector machine discriminant model for the four components of samples was established. The char? acteristic wavelength was selected by the continuous projection algorithm and the second derivative method,and the band with high discrimination accuracy was selected,with a component discrimina? tion rate reached 86. 97%. |
Close |