| Near infrared spectroscopy technology is a hot research topic in noninvasive measurement of blood glucose.Because of its non-contact,sensitive reaction and no need of test paper,the near-infrared spectroscopy has important practical significance for the treatment of diabetes.However,the near-infrared absorption of the sample is usually weak and the spectrum is overlapped.Therefore,choosing the modeling band containing the sample information is the key problem to be solved by near-infrared spectroscopy.In this paper,the near infrared spectral prediction model for the concentration of glucose in human eyes is studied.First in the near infrared spectrum of glucose solution as the test object,establish the basic correction model by partial least square method with the standard deviation of the traditional RMSEP and R correlation coefficient as evaluation index model;to solve the problem of the traditional prediction model of stability and low forecasting precision,this thesis proposes interval combination moving window partial least squares method and improved genetic algorithm select the best band,modeling of glucose solution for 1567nm-1641nm.Then the configuration of the simulation of aqueous solution containing glucose,urea,ascorbic acid and lactic acid salts,collecting its absorbance spectra were established near infrared spectral prediction model using two methods of traditional partial least squares method and proposed,and the prediction accuracy of the model is analyzed.Finally,based on this,the near infrared spectrum analysis system based on Matlab is compiled,which provides a good data processing platform for the researchers. |
PDF Full Text Request