A Rapid And Non-destructive Approach To Identify Bone Fragments Embedded In Pork Based On Multispectral Imaging

Pork products are one of the most popular meat products in China,and its post-processing and food safety have always been one of the issues that people pay close attention to.As one of the unavoidable hazards in raw meat,bone fragments will not only damage the processing equipment,but also cause potential harm to consumers.Therefore,the rapid and non-destructive identification of bone fragments embedded in pork is of positive significance to development of the meat industries.This paper took fresh and frozen-thawed lean pork as research objects.Multispectral imaging were acquired through multispectral images system,by using linear discriminant analysis(LDA),partial least squares discriminant analysis(PLS-DA),and support vector machine(SVM),neural network(NN),successive projection algorithm(SPA),principal component analysis(PCA)and other chemometric modeling methods,which were used for accurate classification of fresh and frozen-thawed pork and rapid and non-destructive identification of bone fragments embedded in pork at the same time.The following conclusions are as follows:1)LDA,PCA-LDA,PLS-DA and SVM chemometric models based on spectral information were established to identify fresh and frozen-thawed pork.The classification accuracy of each model was higher than 86.67%.among which the classification accuracy of,the optimal LDA model based on full spectrum and characteristic spectrum was identified.The accuracy was 100% and 97%,respectively.Model coefficients of spectral information and image information was used to establish a visual distribution map of fresh and frozen-thawed pork,it is possible to identify fresh and frozen-thawed pork,which can identify fresh and freeze-thaw pork more intuitively.2)LDA,PCA-LDA,PLS-DA and SVM metrology models based on spectral information were established to identify the bone fragments embedded in the surface and inside of fresh lean pork.The accuracy of the LDA optimal model based on the characteristic spectrum reached 100%;meanwhile,the image based on the wavelength of the characteristic spectrum could realize image classification and visualization of bone fragments embedded in the fresh lean pork;3)SVM and NN metrology models based on spectral information were establishedto identify the bone fragments embedded in the surface and inside of frozen-thawed pork.The cognitive accuracy of the SVM optimal model based on the characteristic spectrum reached 100%.SVM and NN image classification based on the image information at the full spectrum wavelength to realize the image classification visualization of bone fragments in frozen-thawed pork;The results of this study showed that multispectral imaging technology combined with chemometric modeling methods can be used to achieve accurate classification of fresh and frozen-thawed pork,and at the same time quickly and non-destructively identify the broken bone embedded in pork.The bone fragments embedded in pork could be quickly identified based on multispectral imaging technology,which would provide a theoretical basis for industrial online detection.