Beef Cut Classification Using Multispectral Imaging and Machine Learning Method.

Area of Science:

• * Food science and technology
• * Agricultural analytics
• * Spectroscopy and imaging

Background:

• * Traditional beef cut classification methods are slow and unsuitable for modern industry demands.
• * Multispectral imaging (MSI) offers rapid, non-destructive analysis for food and agricultural products.
• * Accurate beef classification is crucial for food industry applications and product authentication.

Purpose of the Study:

• * To develop a beef cut classification model utilizing multispectral imaging (MSI) and machine learning.
• * To evaluate the effectiveness of single- and multiple-modality feature sets for classification accuracy.
• * To compare the performance of linear discriminant analysis (LDA), support vector machine (SVM), and random forest (RF) classifiers.

Main Methods:

• * Beef samples were captured using a snapshot multi-spectroscopic camera (500-800 nm).
• * Feature sets were engineered using single- and multiple-modality approaches.
• * Machine learning classifiers including LDA, SVM, and RF were employed for model development.
• * Multiple modality feature fusion was investigated to enhance classification performance.

Main Results:

• * The optimized LDA classifier, using multiple modality feature fusion, exceeded 90% prediction accuracy.
• * SVM and RF classifiers also demonstrated satisfying accuracy when combined with machine learning and feature fusion.
• * The study successfully identified effective feature sets and classifiers for beef cut classification.

Conclusions:

• * Machine learning and feature fusion show significant potential for accurate meat classification using multispectral imaging.
• * This approach provides a viable, rapid, and non-destructive alternative to traditional analytical methods.
• * The developed model has implications for future agricultural applications in food analysis and authentication.