The Collection For Near-infrared Spectroscopy Of Aqueous Glucose Solutions And Research On Characteristic Wavelength Optimization Algorithm

Diabetes is one of the most difficult diseases to cure and is associated with a variety of complications.The detection of blood glucose helps the patient's condition control and treatment,but the traditional blood glucose test requires the collection of blood samples,which brings a lot of inconvenience to life.Near infrared spectroscopy is a fast non-destructive testing technology that has received increasing attention in the field of blood glucose testing.However,in the near infrared spectroscopy,there are often a large number of non-information wavelengths or even noise wavelengths,which interfere with the measurement of blood glucose concentration and increase the amount of modeling calculation.Therefore,selecting the wavelength of specific absorption of glucose can effectively filter out noise interference and reduce the calculation.It is a key link in the near infrared spectroscopy analysis.Many scholars have proposed dozens of algorithms.Combining the principle and characteristics of near infrared spectroscopy,this paper builds an experimental platform and obtains near-infrared spectra of 20 samples with different concentrations of glucose solution.Used to find glucose sensitive wavelengths.This paper studies the principle of existing choice of characteristic near infrared wavelengths method.using Anaconda's Spyder software compile the Python programs.The principal component analysis,moving window partial least-squares regression and genetic algorithm are written by the Python programming language.The characteristic wavelengths were searched from the sample data collected,and the partial least squares modeling was performed using the searched wavelengths.The results obtained by the above three methods are better than those of full spectrum modeling,which shows that the optimization of wavelength can improve the modeling effect.In this paper,we optimize on the basis of simple genetic algorithm.The use of elimination and selection two search mechanisms,and sets up multiple sub-population collaborative search,which improve the problem of slower genetic search and avoid easy premature convergence.The improved algorithm achieved the expected effect in the characteristic wavelength search.The characteristic wavelength of glucose is in threeregions,901.44~984.99 nm,1070.61~1178.94 nm and 1372.28~1672.98 nm.The construction of the experimental platform and the preparation of the data processing program provide a convenient spectrum analysis platform for scientific research.