Near-infrared spectrum detection of tobacco nicotine content based on morphological wavelet

In order to improve the accuracy of non-destructive detection of nicotine content of tobacco, a novel method was proposed to get the pretreatment of near-infrared (NIR) spectrum based on morphological wavelet de-noising. The principle and steps of the method were given. The first derivative NIR spectrum of tobacco was served as the target to evaluate the application effect of this method. Then the tobacco nicotine content was calculated based on the de-noised spectrum, and it was compared with the result from wavelet method. Experimental results show that as a kind of nonlinear wavelet, morphological wavelet has both the morphological characteristic of mathematical morphology and the multi-resolution feature of wavelet. It had good performance in keeping details of spectrum and resisting noises. Compared to wavelet threshold method, it had more ideal effects that the correlation ratio (r2) of the prediction set was improved from 0.9877 to 0.9931, and the RMSEP reduced from 0.0539 to 0.0492. The method can be a reference for improving the accuracy of the detection of nicotine content of tobacco and the robustness of model by using near infrared spectroscopy.