Study On Identifucation And Classification Of Clostridium Botulinum Based On Bacteria Electron Microscopic Image And Bacteria Raman Spectra

Bacterial foodborne poisoning has always been a problem that needs attention and is difficult in the field of disease control.Among them,Clostridium botulinum is an important food-borne pathogen.Traditional testing generally takes 5-10 days.It is urgent to propose a new bacterial detection method to meet the needs of today’s society for rapid and convenient bacterial detection.Electron microscopic images have the characteristics of rapid statistical classification and are applicable in many fields.At present,the microscopic image technology still needs to stain the image,and then identify the type of bacteria through manual recognition.The standards are not uniform.Electron microscopic image recognition meets the urgent needs of current bacterial detection by using computers to identify and classify bacteria.As a rapid detection method,spectral analysis avoids unnecessary experimental procedures and provides new ideas for bacterial identification.After extracting the features of bacterial electron microscopy images,SVM,BP neural network,Adaboost algorithm and its improved method,the information entropy-based Adaboost method is used to classify the images.The classification accuracy of SVM is 82.2%;the classification result of BP neural network is 88.71%.The Adaboost series methods show good classification results.The accuracy of them can reach 100%.However,after reducing the number of samples of train set and reducing the number of base classifiers,the information entropy-based Adaboost method can still achieve 100%accuracy,while the original Adaboost algorithm has an accuracy rate of 97.4%.The introduced information entropy is used to measure the"divergence" of the predictions of different base classifiers for the same sample.Samples with large entropy values contain more important information,which is more conducive to the construction of the overall classifier.Alexnet network is also used in the classification of bacterial electron microscopic images,with an accuracy rate of 92.55%.The Alexnet model has good generalization ability and is suitable for identification and classification of bacteria electron microscopic images.In the subject of bacterial Raman spectroscopy classification,the traditional SIMCA and improved SIMCA-SVDD methods are used to classify the spectral data.The accuracy of the original SIMCA algorithm is 96.36%,and the accuracy of the improved SIMCA-SVDD method reaches 98.18%.The improved SIMCA-SVDD algorithm uses the characteristics of SVDD to map low-dimensional linear inseparable feature data to a higher-dimensional space,and then classify the data.Smaller areas can be drawn to contain more class samples to achieve better classification results.