Research on the Authenticity of Mutton Based on Machine Vision Technology

Area of Science:

• Food Science
• Computer Science
• Artificial Intelligence

Background:

• Mutton adulteration is a significant concern for consumers and the food industry.
• Accurate and rapid identification methods are needed to ensure food authenticity and safety.

Purpose of the Study:

• To develop a real-time automatic identification model for adulterated minced mutton using convolutional neural network (CNN) image recognition.
• To evaluate the effectiveness of various CNN models for recognizing meat species and adulterated mutton.

Main Methods:

• Acquired images of pure mutton and mutton adulterated with duck, pork, and chicken using a self-built image acquisition system.
• Compared the performance of six CNN models (AlexNet, GoogLeNet, ResNet-18, DarkNet-19, SqueezeNet, VGG-16) for image recognition.
• Utilized a dataset including 960 images of different animal species, 1200 images of adulterated minced mutton, and 300 images for external validation.

Main Results:

• ResNet-18, GoogLeNet, and DarkNet-19 demonstrated the best learning effects and accuracy in identifying meat pieces and adulterated mutton.
• Training accuracy for these top models exceeded 94%.
• External validation accuracy for identifying adulterated minced mutton reached over 70%.

Conclusions:

• Deep convolutional neural network (DCNN) models can effectively identify different livestock and poultry meat pieces and adulterated mutton.
• This image learning approach provides technical support for rapid, non-destructive identification of mutton authenticity.