The Identification Of Herbage Varieties Using Terahertz Spectra

In this study,terahertz time-domain spectroscopy were explored to demonstrate the spectral characteristic of the alfalfa forage seeds.Combined with Statistical analysis methods,all kinds of spectra were used to classify the variety of these seeds.Based on these spectrum,the methods of optimizing data and establishing database were explored.And finally,using the chemical composition to reveal the corresponding base principle.1.Forage seeds were detected by Transmission type terahertz spectrum system.In the effective frequency range,there are differences in the absorption coefficient and the refractive index of each sample.Although there are differences,those distinctions can not be the classified standard so that these spectrum were explored which combined with Statistical analysis methods to analysis deeper.2.First,partial least squares(PLS)is used to establish model of six kinds of sample which are caoyuan No.2?caoyuan No.3?gannong No.3?gannong No.4?gannong No.7and gannong No.8 from the same year.The result shows that the data of the refractive index have higher precision of prediction,then cluster analysis(CA)is used to calculate the data and it is successful to classify the two series.Though these classification can not be seem as the classified standard too.3.In order to classify the forage seeds better,this paper uses three kinds of PLS to calculate their best frequency bands respective,and test the effect of clustering on the base of the data from these frequency bands.The results show that the higher precision of prediction relates to the the better effect of clustering.4.this paper combined the biological characteristics with the principal component analysis(PCA)to explore the basic theory about detecting material by the terahertz time-domain technology.Comparing the result of PCA with the chemistry composition and references,the protein have been turn out to be play a leading role,and other composition have a certain impact on the difference of alfalfa seeds.5.In order to build a database of alfalfa forage seeds,two kinds of discriminant analysis were used to establish model of database and predict the unknown samples.Comparing two methods,the results showed that these two methods have the effectiveand same results about classifying the sample,and the accuracy reached 87.5%,which turn out the method discriminant analysis can be used to establish the database based on the terahertz spectrum.These consequences proved that using Statistical analysis methods can be a workable direction of establishing a alfalfa forage database.