Determination of animal skeletal maturity by image processing

Area of Science:

• Agricultural Science
• Computer Vision
• Food Science

Background:

• Accurate beef carcass maturity assessment is crucial for quality grading.
• Traditional methods rely on subjective evaluation of cartilage ossification.
• Objective, automated methods are needed to improve consistency and efficiency.

Purpose of the Study:

• To develop and evaluate a computer vision system for predicting USDA beef maturity grades.
• To investigate the utility of color image features, specifically hue, for characterizing skeletal maturity.
• To assess the performance of a neural network trained on these features.

Main Methods:

• Color image features (RGB, HSL) were extracted from thoracic vertebrae cartilage.
• A feature curve based on hue value was defined to represent color variations.
• A neural network was trained using hue-based feature curves to predict maturity grades.
• Samples from two processing plants, with different grade distributions, were used for testing.

Main Results:

• Hue value was identified as the most effective color feature for maturity assessment.
• Mean hue values of cartilage showed significant differences (P<0.05) across maturity grades.
• The neural network achieved prediction accuracies of 75% and 65.9% on the two sample sets.

Conclusions:

• Computer vision techniques show significant potential for objective beef maturity assessment.
• Color-based image analysis, particularly hue, can effectively characterize skeletal maturity.
• This approach offers a promising automated alternative to traditional beef grading methods.