Automated MSA/AusMeat hyperspectral handheld grading for beef

Summary

The MLA funded project (P.PSH.0776) entitled "Automated MSA/AUS-MEAT hyperspectral handheld grading for beef" was initiated in 2016 and finishes at the end of 2019. The project has been conducted in collaboration with Murdoch University (Perth) and University of New England (Armidale) and has strong links to the Advanced Livestock Measurement Technologies (ALMTech) program.

The main objective of the project was to develop an automated handheld camera which grades the beef rib eye according to MSA/AUS-MEAT standards consistently and in accordance with visual grading conducted by highly trained graders. The overall project goal was to make available an instrument with prediction models that will enable all abattoirs in Australia to grade carcases in a uniform and consistent way and thereby creating high value for the Australian beef industry.

Vision-based systems for grading beef rib eye already exist on the market. However, the commercially available equipment does not appear suitable for all Australian beef grading. In the present project a lamb grading based handheld instrument has been further developed into one suitable for the beef industry which allows automated classification of physical attributes such as ribeye muscle area, marbling, subcutaneous fat depth and colour of fat and meat tissue. Traits which up until now are being visually graded by accredited AUS-MEAT and MSA graders according to AUS-MEAT's Chiller Assessment Language. Three handheld camera prototypes were built and used for testing by Murdoch University.

A total of 888 steers were included in the project and samples were collected under commercial slaughter conditions from steers belonging to the Northern and Angus Beef Information Nucleus (BIN) herds. Carcasses were quartered between the 12th and 13th rib and allowed to bloom for at least 30 minutes. Care was taken to ensure that bone dust and fat was removed from the surface of the ribeye area. Carcasses were then graded according to the AUS-MEAT Chiller Assessment Language by a Meat Standards Australia (MSA) grader between 8 – 24h after slaughter for all relevant carcass traits. The evaluated carcass traits being: ribeye area, subcutaneous rib fat thickness, MSA marbling score, AUS-MEAT marbling score, and meat and fat colour class. Subsequently, images were captured with the handheld camera equipment developed by Frontmatec Smoerum A/S. A sample was collected for NIR-based determination of the intramuscular fat content by Murdoch University.

In conclusion, algorithms for ribeye area, intramuscular fat content, marbling score against both MSA and AUS-MEAT classification and meat colour have been developed. Development of an algorithm for fat colour classification awaits a data set which covers a larger spread in colour classes with each class properly represented. Subcutaneous rib fat thickness cannot be determined by the camera solution. An alternative to subcutaneous rib fat would be total fat thickness. This trait is feasible but not part of the AUS-MEAT standard and reference data for this trait was not obtained in the project.

Based on the learnings from the current project Frontmatec intends to develop a commercial device after completion of P.PSH.0776.