Analysis of beef quality according to color changes using computer vision and white-box machine learning techniques.

Area of Science:

• Food science and technology
• Computer vision
• Machine learning

Background:

• Beef quality is intrinsically linked to its color, with a cherry red hue indicating freshness.
• Brownish tones signify quality degradation.
• Current methods often analyze color channels independently, limiting comprehensive assessment.

Purpose of the Study:

• To develop and evaluate a multivariate approach for beef color analysis using white-box machine learning.
• To compare the effectiveness of multivariate analysis against traditional single-channel methods for predicting beef freshness.
• To identify optimal color spaces and machine learning models for accurate beef color prediction.

Main Methods:

• Utilized a Computer Vision System (CVS) with a color correction cabin for objective beef color capture.
• Investigated three color spaces: RGB, HSV, and CIELab*.
• Evaluated three white-box classifiers: decision tree, logistic regression, and multivariate normal distributions for predicting beef freshness.

Main Results:

• The multivariate analysis models demonstrated high accuracy in predicting beef color for both fresh and non-fresh samples.
• Significant differences were observed in the performance across the evaluated color spaces.
• White-box models provided a clear understanding of the prediction process.

Conclusions:

• Multivariate analysis of beef color offers superior prediction accuracy compared to analyzing individual color channels.
• The proposed machine learning approach enhances objective assessment of beef quality and freshness.
• This method provides a robust framework for quality control in the beef industry.