Tobacco substitution and cigarette blend maintenance based on near infrared spectral similarity

Near infrared spectroscopic spectral similarity was investigated and applied to tobacco substitution and cigarette blend maintenance. By means of a local spectral pre-processing and an ensemble spectral similarity algorithm, any interference on baseline and scattering in the spectrum was eliminated and the technique was used for characterizing tobacco leaf based on spectral similarity to aid cigarette blending. The results showed that the matching rates according to the spectral similarity with target strips were 82.0% and 75.7% viewed from the producing area and stalk position respectively. The average differences on main chemical components, such as total sugar and nicotine, were less than 5%. There was no significant difference in chemical components and sensory indexes between the original blend and a simulated blend obtained through combining strip similarity with strip combination similarity.