Detection Of Bacterial Microbes In Milk By Two-dimensional UV-Vis Spectroscopy

Microbial contamination in milk is one of the main factors of the dairy industry.At present,plate counting method is still the main method for the detection of microorganism in dairy products,but the samples need to continuous acquisition,experimental operation process complicated and time-consuming.Therefore,the rapid,accurate and low-cost microbiological detection technology based on milk infection microorganisms is of great significance to the dairy detecting industry..UV-Visible spectroscopy was used in this study.Escherichia coli,which is often detected in milk,and Staphylococcus aureus with similar characteristic absorption curve were used as bacterial types,and Saccharomyces,which is easily infected in milk,was used as detection object.Milk was used as detection background to study the concentration detection method of these three microorganisms in milk infected with microorganism.The specific work is as follows:（1）Different bacteria microbes in water and milk UV-Vis spectral absortion characterial analysis: Through the conjugate structure of the microbial material suction light,substituent effect and the solvent effect of milk three point of view,analysis for the cause the characteristic absorption curve of milk and microorganism,for the subsequent testing experiment design to provide theoretical foundation and basis.（2）Design the u V-vis spectrum acquisition experiment of the microorganism to be tested in milk,and extract the characteristic wavelength representing the bacterial microorganism in milk.According to the principle of two-dimensional correlation analysis,the dynamic spectral data matrix of microbial concentration gradient was acquired.Through two-dimensional correlation analysis method,the extension of the one dimensional spectrum to two-dimensional,treat survey of microbial characteristics to absorb the pick up point,solve the milk spectral data of microbial spectrum data cover problem,at the same time using the pick up to a representative of the microbial concentration changes of feature point concentration prediction model is set up,avoid full spectrum data input model of fitting problem.Before establishing the model,a combination of spectral pretreatment methods was used to preprocess the spectral data.According to the comparison of processing results,an appropriate spectral pretreatment method was selected to solve the scattering problem of milk as a solvent.（3）A prediction model of microbial concentration in milk was established.According to the characteristic bands extracted from two-dimensional correlation analysis,the partial least squares regression（PLSR）model and support vector machine regression（SVR）model were established respectively,and the least squares support vector machine regression（LS-SVR）model was established.By comparison,the analysis results show that the least square support vector machine has the fastest fitting speed and better effect,and the overall prediction accuracy of the model is93.57%,which meets the needs of rapid detection in this topic.