Study On The Detection Of Classification Of Meat Tissue By Laser-induced Breakdown Spectroscopy

The classification of meat tissues is of great significance for the standardization of meat markets and the diagnosis of pathological tissues.Traditional meat tissue classification is mainly based on DNA and protein detection.It has high accuracy,but the operation is complicated and time-consuming.Therefore,how to find a simple and rapid classification method of meat tissue has become a research hotspot in recent years.Laser-induced breakdown spectroscopy,which is fast,in-situ,micro-destructive,remote-detection and easy to operate,has great application prospects in the field of meat tissue detection.However,many problems restrict the further development of LIBS technology in the classification of meat tissues,such as complex sample preparation,spectral information redundancy and noise interference.Therefore,this work aims to explore sample preparation methods and classification algorithms based on feature extraction.Common meat tissues and Graves' ophthalmopathy disease tissues were detected in this work.The main achievements and innovations are as follows:(1)We first applied slice preparation methods such as frozen sectioning,paraffin sectioning and vibration sectioning in the field of LIBS.Slice preparation methods can make the surface of meat tissue samples flat and uniform thus improve the spectral quality.It is also suitable for small-sized meat tissue samples.Therefore,slice preparation benefits for expanding the scope of application of LIBS meat tissue detection.We selected common meat tissues such as black goat mutton,chicken breast,pork tenderloin and beef tenderloin,and pretreated them by the slice preparation methods.The results of comparative studies on spectral quality and classification showed that the frozen section method and the paraffin section method improved stability of spectrum and universal,the two methods were recommended for LIBS meat tissue detection.(2)This work first introduced spectral feature extraction methods such as minimum noise separation(MNF),isometric feature mapping(ISOMAP)and linear discriminant analysis(LDA)to remove redundant information of excessive spectral lines and noise of LIBS spectrum.And then linear discriminant analysis(LDA),K nearest neighbor algorithm(KNN),extreme learning machine(ELM),support vector machine(SVM)and Bagging classification algorithm were employed to classify the tissues.The experimental results showed that the SVM performs the best with prediction accuracy of 88.45% when using the classification algorithm alone.When combining the classification algorithm with spectral feature extraction algorithm,the prediction accuracy was improved and classification effect combined with SVM was also the best.The prediction accuracy of MNF-SVM,ISOMAPSVM,and LDA-SVM classification models were 97.05%,97.30%,and 99.93%,respectively.Considering the simple,efficient and unsupervised characteristics of the MNF algorithm,MNF-SVM was the best classification model in meat tissue.(3)The paraffin section method and the MNF feature extraction algorithm were applied to the detection of Graves' ophthalmopathy pathological tissue in LIBS.The MNF feature extraction algorithm was combined with LDA,KNN,and SVM classification algorithms.After comparing the classification results when using the classification algorithm alone,the appropriate classification algorithm KNN and SVM were selected.MNF was used to extract features on the spectrum then KNN and SVM were employed to classify separately.The results showed that the prediction accuracy,sensitivity,specificity,and recall of MNF-SVM were 99.39%,98.78%,100%,and 100%,respectively,and the coefficient of variation was 1.76%.All indicators were better than MNF-KNN.In summary,MNF-SVM algorithm combined with the slice preparation methods proposed in this work can effectively overcome the problems of uneven surface,poor spectral quality and spectral noise interference caused by loose samples using LIBS in meat tissue detection.The classification results show that LIBS technology combined with the appropriate feature extraction classification algorithms has broad application prospects in the detection and classification of meat tissues and pathological tissues.