Journal Article

Robust and Accurate Classification of Mutton Adulteration Under Food Additives Effect Based on Multi-Part Depth Fusion Features and Optimized Support Vector Machine

Huang, Zhongtao, Cui, Xiaomin, Zhu, Rongguang, Li, Bin, Yao, Xuedong, Wang, Shichang

2023

Overview

The adulteration of pork mixed in mutton is pervasive in the market. However, when the adulterated mutton with multi-part pork is mixed, the discrimination becomes difficult due to differences in composition. Meanwhile, with the diversification of adulteration methods, food additives are also mixed into adulterated mutton to interfere with discrimination. In this study, the discrimination of mutton adulteration with multi-part pork (from back, hind leg, and front leg) under the influence of mutton flavour essence and colourant was explored using NIR-HSI. A novel framework in which three parallel convolutional neural networks (CNNs) serve as feature extractors was designed to obtain the multi-part depth fusion features of the sample. After obtaining fusion features, classification models were established by using back propagation neural network (BPNN), random forest (RF), and support vector machine (SVM). Sparrow search algorithm (SSA), genetic algorithm (GA), and particle swarm optimization (PSO) were employed for parameter optimization of classifiers. The results showed that the performance of models based on fusion features was significantly better than that of models without considering the characteristics of pork parts. The optimized SVM classifier via SSA obtained the best result. The overall accuracy, F1-score, and kappa value of the external validation set were 98.61%, 97.86%, and 96.61%, respectively. Overall, NIR-HSI combined with CNN and optimized SVM could function as a robust and accurate detection method for discriminating adulterated mutton with multi-part pork under food additives effect.

• Publisher: Springer US,Springer Nature B.V
• Subject: adulterated products / Algorithms / Analytical Chemistry / Artificial neural networks / Back propagation networks / Chemistry / Chemistry and Materials Science / Chemistry/Food Science / Classification / Deep learning / Feature extraction / flavor / Food additives / Food Science / Genetic algorithms / Machine learning / markets / Meat / Microbiology / Mutton / Neural networks / Particle swarm optimization / Pork / Robustness / Search algorithms / Solvents / Support vector machines