Skip to main content
ARS Home » Midwest Area » Wooster, Ohio » Corn, Soybean and Wheat Quality Research » Research » Publications at this Location » Publication #371112

Research Project: Genetic and Biochemical Basis of Soft Winter Wheat End-Use Quality

Location: Corn, Soybean and Wheat Quality Research

Title: Machine learning in the assessment of meat quality

item Penning, Bryan
item Snelling, Warren
item Woodward-Greene, Jennifer

Submitted to: IEEE IT Professional
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/6/2020
Publication Date: 5/22/2020
Citation: Penning, B., Snelling, W.M., Woodward Greene, M.J. 2020. Machine learning in the assessment of meat quality. IEEE IT Professional. 22(3):39-41.

Interpretive Summary: Currently, the quality of beef moving through a production line is graded by a trained USDA inspector for both potential quality and quantity of meat with just 6 seconds to make a determination. This paper describes efforts by USDA researchers to use new technologies such as imaging systems and specialized mass spectrometry paired with computer machine learning and artificial intelligence techniques to estimate meat quality and quantity with greater accuracy and high speed. For quantity and some quality traits, machine learning and artificial intelligence systems when properly calibrated were able to approach or exceed the accuracy of the USDA inspector. These technologies paired with trained USDA inspectors may improve the speed and accuracy of meat quality and quantity estimates benefiting both the beef industry and consumer.

Technical Abstract: We compare two approaches to automate carcass quality grading using different artificial intelligence methods. The first is based on image analysis, and the second uses state-of-the-art Rapid Evaporative Ionization Mass Spectrometry (REIMS). Both employ machine learning (ML) to increase the speed and accuracy of carcass quality evaluation. The image analysis method increased speed and accuracy for all quality measures except marbling when compared to human meat inspectors. The mass spectrometry method tested eight ML algorithms, and achieved an impressive 81.5 to 99% accuracy in predicting carcass quality traits. However, this accuracy was dependent on the trait examined, so ML algorithms were not the answer for all traits.