White striping degree assessment using computer vision system and consumer acceptance test.

Area of Science:

• Animal Science
• Food Science
• Computer Science

Background:

• White striping (WS) is a myopathy affecting broiler breast meat quality.
• Objective assessment of WS severity is crucial for the poultry industry.
• Consumer perception of WS impacts purchasing decisions.

Purpose of the Study:

• To evaluate three degrees of white striping (WS) for automatic assessment using a computer vision system (CVS).
• To explore machine learning (ML) algorithms and image features for WS classification.
• To assess consumer acceptance and purchase intent related to WS severity.

Main Methods:

• Trained specialists classified WS severity based on visual and firmness aspects.
• A digital camera captured images, from which 25 features were extracted.
• Machine learning algorithms (SVM, fuzzy-W, Random Forest) were applied for classification; sensory analyses evaluated consumer acceptance and purchase intention.

Main Results:

• Classification models achieved up to 86.4% accuracy, with Support Vector Machine, Fuzzy-W, and Random Forest performing best.
• Multilayer Perceptron showed lower accuracy (70.9%), particularly with normal samples.
• Sensory analysis revealed WS negatively affects grilled fillet texture and raw sample appearance, influencing purchase intent.

Conclusions:

• The developed CVS is accurate and fast for classifying WS samples.
• WS myopathy negatively impacts broiler breast meat tenderness and visual appeal.
• The findings highlight the importance of WS in consumer acceptance and purchase decisions.