Pollen Classification Based On Raman Spectroscopy

Pollen is the germ cell in the stamen of seed plants.In recent years,the number of people with pollen allergy has been increasing rapidly,and pollen allergy has become a major threat to public health.Pollen allergens and allergic reactions vary from person to person.How to quickly detect the classification of pollen particles in the air is a very practical research topic.Raman spectroscopy is a detection technology based on the inelastic scattered light of samples,which has the advantages of fast,nondestructive and high sensitivity.In recent years,it has been widely used in biomedical,food safety,environmental monitoring and other fields.In this paper,a large number of pollens were collected by Raman spectrum and a spectral library was established.Based on the biological fingerprint of pollens Raman spectrum,an identification model was established and the inter-family and inter-species pollens were identified and classified.This subject mainly carries out research work from the following aspects:1.collect Raman spectra of pollen samples.Raman spectrograph was used to collect 178 kinds of pollens from different seasons.The comparative analysis of experimental parameters showed that the excitation wavelength was 633nm,the integral time was 20S,the confocal aperture was 200?m,the scanning range was300cm^-1-2000cm^-1,and the spectral resolution was about 1.5cm^-1.It is worth noting that when the pollens are used as biological samples,the fluorescence background is strong,so the samples should be irradiated and quenched by photobleaching before the experiment.2.sample Raman spectral pretreatment and analysis.Firstly,through comparative analysis,the Savitzky-Golay method with a window size of 5 points was selected to smooth the original spectrum.Secondly,the linear fitting method is used to remove the baseline.Finally,the spectrum is normalized.Pretreatment can solve the problems of noise,baseline drift and order of magnitude difference as far as possible.Multiple spectral data from the same sample should be averaged to reduce the random error caused by the specificity of different individuals.3.establish identification model.According to the biological classification,178kinds of pollen samples after pretreatment were classified into 28 kinds of inter-family samples.Each inter-family sample was collected for principal component the results.The classification model was obtained by training the training set composed of 190 samples by support vector machine.The classification model was used to predict the prediction set composed of 89 samples,and the result showed that the comprehensive discrimination accuracy was 97.75%.Furthermore,the identification model was established by selecting rosaceae,rhinoceros and compositae from the samples of the same family.The accuracy of 6 rosaceae SVM recognition models was 90.47%,and that of 7 rosaceae models was 85.18%.The above results indicate that Raman spectrum can be used to distinguish the samples of multiple species in genus.In addition,the effects of carotenoids on Raman spectra of pollens were described by using the special properties of compositae samples.Through the above research,this paper established Raman spectral library and an effective classification model of pollens,proved the feasibility of using Raman spectral technology to classify and identify pollens,and opened up a new path for automatic detection and early warning of pollens.