A Study Of Blood Species Identification Method And Instrument Realization Based On Spectral Analysis

The blood sample,as a kind of special products,contains unique genetic resources of organisms.If this part of the resources is mastered by someone with ulterior motives,targeted individualized drugs for species in our country may be developed through its inclusion of biological genes,and thus achieve the advantages in the field of medicine of our country,or threaten the health and safety of national life.At present,the species of blood can be identified by observing cell morphology,flow cytometry,or DNA detection,etc.All of above methods are contact-type detection methods,and therefore the samples may occur cross-contamination during the test process.Hence,the development of non-contact blood species identification method has great significance.In this paper,the generation mechanism of supercontinuum source is quested,and the method of blood species identification based on spectroscopy is studied deeply.A preprocessing method is proposed,which applying the genetic algorithm to the original spectral data in order to obtain the contribution of different bands to the classification,and weighting the raw data according to the contribution.Then classify the preprocessed spectral data through ensemble learning method,and get no less than 90% spectral classification accuracy.The validity of the algorithm is verified by a total of 958 samples from 17 species in this paper.At the end of the paper,the design of non-contact blood species identification instrument was introduced.The paper has following innovation points:1.Classify the species of blood samples by using machine learning methods including k nearst neighbors algorithms,support vector machines,na?ve Bayesian classification,and ensemble learning methods.2.Proposed a method which uses the genetic algorithm to obtain the weight of different bands' contribution to classification,and a function to evaluate the contribution of classification.The method is validated by known data and the expected results are obtained.3.For the first time completed the design of the non-contact blood species iden-tification instrument,and achieved a classification accuracy of not less than90%.4.Built up the database which contains a variety of species of whole blood,serum and plasma diffuse absorbation and fluorescence spectrum.