In-line image analysis in the slaughter industry, illustrated by Beef Carcass Classification.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Meat Science

Background:

• Quantitative quality control of biological objects is essential for industry standards.
• The Beef Carcass Classification centre (BCC-2) was developed as a second-generation prototype for automated carcass analysis.
• Existing methods lacked consistent calibration and advanced processing for biological variations.

Purpose of the Study:

• To describe a complete computer vision framework for quantitative quality control of biological objects.
• To detail the implementation and capabilities of the BCC-2 system for beef carcass analysis.
• To establish a procedure for maintaining calibration across multiple BCC-2 units over time.

Main Methods:

• Utilized computer vision techniques, including traditional pattern recognition, principal component analysis, and neural networks.
• Employed a system (BCC-2) with a frame, camera, PCs, terminal, and slide projectors for 3D shape analysis.
• Measured carcass properties such as conformation, fatness, fat color, saleable meat percentage, and rib eye cross-sectional area.

Main Results:

• BCC-2 quantitatively measures beef carcass geometry and color.
• The system successfully determines visual properties and objective quantities like saleable meat percentage.
• A calibration procedure ensures consistent performance across multiple BCC-2 units.

Conclusions:

• The developed framework provides a robust and adaptive solution for automated beef carcass classification.
• BCC-2 demonstrates the effectiveness of advanced information processing for handling biological variations in carcasses.
• The system is cost-effective, built from inexpensive components, and robust in classifying carcasses.