Real-time Sorting of Broiler Chicken Meat with Robotic arm: XAI-enhanced Deep Learning and LIME Framework for Freshness Detection

Abstract

Ensuring food safety in poultry processing is crucial due to the risk of contamination, which can lead to health hazards. Traditional methods for sorting chicken meat based on freshness are often manual and prone to error, necessitating the development of an automated, accurate sorting system. This research focuses on integrating Convolutional Neural Networks (CNNs) with Local Interpretable Model-agnostic Explanations (LIME) to enhance the accuracy and efficiency of sorting broiler chicken meat into fresh and rotten categories. This study employed the InceptionV3 CNN model combined with LIME for transparent decision-making. The model was trained and tested on a dataset of chicken meat images to guide a robotic arm for real-time sorting. The system achieved an impressive sorting accuracy of 96.3 %, outperforming traditional manual methods. With the guidance of the CNN-LIME integration, the robotic arm successfully processed 1000 samples of fresh and 300 samples of rotten chicken meat, achieving a precision of 94.19% for fresh meat and 97.24% for rotten meat. This integration significantly improved sorting accuracy and efficiency in poultry processing, providing a reliable method for distinguishing between fresh and rotten chicken meat. This system improves food safety and operational efficiency in automated food processing. It shows the effectiveness of combining explainable AI with robotics, reducing human exposure to harmful contaminants ethically. It also contributes to public health by ensuring safer food products. This research presents a new approach to automated food sorting by integrating CNNs with LIME for transparency and accuracy. This can be beneficial for food processing industries and researchers in the field of AI and robotics.

Keywords

• Food safety
• Poultry processing
• Chicken meat sorting
• CNN
• Explainable AI
• LIME