Rapid identification of water-injected pork and gum-injected pork by the combination of low field nuclear magnetic resonance and chemometrics analysis

Abstract

ObjectiveTo apply rapid identification of water-injected and gum-injected pork by chemometrics analysis combined with low field nuclear magnetic resonance.Methods17 variables obtained by low field nuclear magnetic resonance were used to extract principal components by principal component analysis， and 4 principal components were identified by discriminate analysis. A three-layer MLP neural network model with a structure of 17-6-3 was established through constrained optimization modeling using 17 original variables as input， and the types of water-injected pork and gum-injected pork as output.ResultsIn the discriminate analysis model， the discriminant accuracy of cross-validation in the control group， water-injected group and gum-injected group was 60/60， 91/95 and 359/384， and the total correct discriminant rate was 94.6%. In the neural network model， the total correct discriminant rate of 179 samples in the testing set was 97.8%.ConclusionLow-field nuclear magnetic resonance combined with discriminate analysis model and neural network model can effectively distinguish water-injected and gum-injected pork. The method is simple and can be used for rapid detection.

Keywords

• water-injected pork
• gum-injected pork
• low-field nuclear magnetic resonance
• principal component analysis
• discriminate analysis
• multi-layer perceptron neural network
• rapid detection