The Study Of The Processing And Diagnosis Of The Infrared Spectrum Of Signal In The Classification Of Tobacco Leaves

The neural network applied in automatic grading of tobacco leaves based on the infrared spectrum is studied.The absorbed spectrums of tobacco leaf are described by the infrared because its main compositions are carbon, hydrogen groups (such as C-O, C-H groups). That is the infrared spectrum of tobacco leaf is closely related to the inner characteristics and chemical composition. Therefore, this paper try to grade and group the tobacco leaves based on their infrared spectrum by neural network.In Current, the tobacco grading in our country is still remained in the stage of artificial classification. This artificial hierarchical method need is a wealth of experiences. Obviously, there are strong subjective and arbitrary. The studies on the intelligence grading of tobacco have focused on the processing of images. They mostly extract image features and grade the tobacco leaves by the method of the nest neighbors. However, some characteristics which are closely related to the tobacco grade can not be expressed by images, such as thickness, leaf structure and so on. However, they can be well described by their spectrum.Infrared spectrum is widely used in the qualitative and quantitative analyzing the compositions of tobacco leaves. The damaged processing has done when the spectrum are obtained in these studies. This approach can not keep the thickness information of tobacco leaves. Moreover, the grading of tobacco leaves during purchasing must be non-destructive obtaining spectrum signal. The experiments show that the complicated pro-processing for the spectrum is not necessary by compared. And simple preprocessing can make the grading faster with almost same correct classification rate.The probabilistic neural networks, supporting vector and radial basis function neural network were used to group or grade the tobacco leaves based on their infrared absorption spectrum. The results of the experiment showed that the correct rate of the group and grade are above 90%.