Artificial Intelligence (AI) for Airports: Baggage Detection and Size Estimation for Enhancing Operational Efficiency at Small and Medium-Sized Airports

The rapid expansion of global air travel has heightened the demand for automated, intelligence-driven baggage management systems capable of operating in small and medium-sized airports. This paper presents two complementary innovations: (1) a real-time baggage detection and size-estimation pipeline employing the state-of-the-art YOLOv8 deep neural network, and (2) a generative smart-luggage design framework powered by a hybrid genetic-CNN optimization. In the first task, we construct a diverse 2,000+–image dataset annotated via Roboflow and train both YOLOv8n (nano) and YOLOv8m (medium) models to classify luggage by material (e.g., hard plastic, metal, textile) and estimate physical dimensions through calibrated pixel-to-centimeter conversion. YOLOv8m achieves a mean Average Precision (mAP@0.5) of 0.805 and a size-estimation error below 8 mm, demonstrating a robust accuracy–speed tradeoff suitable for deployment on edge-compute platforms. In the second task, we encode luggage attributes (size, weight, functionality, and CNN-derived aesthetic scores) into a multi-objective genetic-algorithm fitness function. Over ten generations, our GA–CNN hybrid converges to novel luggage prototypes that balance ergonomic form factors and smart features (e.g., RFID integration). Experimental results confirm that our approach outperforms prior YOLOv5-based detectors by 12% in detection precision and yields ergonomic designs validated via user-preference surveys. Collectively, these two contributions forge a path toward fully integrated, AI-driven baggage workflows—spanning detection, handling, and personalized product design—paving the way for enhanced operational efficiency and user satisfaction in next-generation airport environment