Studies On Classification Of Bacteria And Determination Of Bacterial Concentration Based On NIR And MIR Spectroscopy

The rapid,accurate and effective detection of microorganisms is of great practical significance for ensuring the quality of life and safety of people,and maintaining social security and stability.In this thesis,Escherichia coli,Staphylococcus aureus,and Salmonella in liquid culture at different concentrations were used as research objects,and near-infrared spectroscopy and mid-infrared spectroscopy techniques were used to classify bacterial species and to deteremine the bacterial concentrations.The main findings are as follows:?1?The optimal bacterial classification model based on near-infrared and mid-infrared spectroscopy was determined.The partial least square discriminant analysis?PLS-DA?,particle swarm optimization support vector classification?PSO-SVC?and genetic algorithm optimized support vector classification?GA-SVC?established by near-infrared and mid-infrared spectral data under different preprocessing are were compared.The research results showed that the performance of the models based on mid-infrared spectroscopy was generally better than near-infrared spectroscopy,and the optimal model was the PLS-DA model based on mid-infrared spectroscopy,where the classification accuracy for calibration and prediction were both 100%.?2?The optimal simplified model for classification of bacteria based on near-infrared and mid-infrared spectroscopy was determined.The successive projection algorithm?SPA?,competitive adaptive reweighted algorithm?CARS?,and the combination of CARS and SPA were used to select the characteristic wavelengths of the near-infrared and mid-infrared spectral,respectively,based on which simplified models were establish.Among them,the accuracy of the simplified model based on the mid-infrared spectrum was higher than that of the model based on near-infrared spectrum,and the GA-SVC model based on the mid-infrared wavelengths selected by using CARS was found to be the best.The classification accuracy for calibration and prediction of the model were 100%and 95.83%,respectively.?3?The optimal bacterial concentration detection model based on near-infrared and mid-infrared spectroscopy was determined.The performance of partial least squares regression?PLSR?models,principal component regression?PCR?models and support vector regression models was compared under different preprocessing of near-infrared and mid-infrared spectral data Coyote optimization algorithm?COA?to proposed to optimize the penalty coefficient C and kernel function parameter g of the SVR,and the result was compared against that based on particle swarm algorithm optimization.The research results showed that near-infrared spectroscopy was not satisfactory for bacterial concentrationdetermination,and the mid-infrared models output better prediction performance.Among them,the COA-SVR model was the best based on the original data.The²and RMSEP of the model reached 0.94 and 0.44 log CFU/m L,respectively.?4?The optimal simplified model for bacterial concentration determination based on near-infrared spectroscopy and mid-infrared spectroscopy.Successive projection algorithm?SPA?,competitive adaptive reweighted algorithm?CARS?and their combination were used to extract features from the near-infrared and mid-infrared spectral data,respectively,based on which simplified models were established.According to the model results,the performance of the CARS-SPA-PLSR model based on mid-infrared spectroscopy was the best with²of 0.87 and RMSEP of 0.60 log CFU/m L,respectively.The results of this study indicate that mid-infrared spectroscopy outperformed near-infrared spectroscopy in both bacterial classification and concentration determination,and mid-infrared spectroscopy can quickly and non-destructively detect the type and concentration of bacterial samples in liquid culture.