Improving traceability and quality control in the red-meat industry through computer vision-driven physical meat...

Area of Science:

• Food Science
• Computer Vision
• Agricultural Technology

Background:

• Current beef traceability systems are vulnerable to tampering due to reliance on external markers.
• Objective quality assessment in beef is challenging with existing methods.
• Unique intramuscular fat patterns offer potential for inherent identification and quality evaluation.

Purpose of the Study:

• To investigate the use of intramuscular fat patterns for beef traceability and quality assessment.
• To develop and validate a machine learning model for analyzing these internal meat features.
• To enhance fraud resistance and objectivity in the red meat supply chain.

Main Methods:

• Development of a large dataset of 38,528 high-resolution beef images from 602 steaks.
• Annotation of images with human grading and ingredient analysis data.
• Implementation of an EfficientNet-based model for quality prediction and traceability analysis.

Main Results:

• High accuracy in marbling score prediction (96.24% top-1±1, 99.57% top-1±2).
• Accurate breed identification (91.23%) and diet determination (90.90%).
• Excellent performance in traceability tasks, with F-1 scores of 0.9942 (sample-to-sample) and 0.9479 (sample-to-database).

Conclusions:

• Intramuscular fat patterns serve as reliable natural identifiers for beef traceability.
• This AI-driven approach enables objective quality assessment and significantly improves fraud resistance.
• Internal meat features offer a robust solution for enhancing integrity in the red meat supply chain.