<![CDATA[The American Statistical Association reports that Bangkok, the capital and largest city of Thailand, holds the top spot as the most visited city worldwide in 2023. X-ray imaging for security screening plays a crucial role in upholding transportation security by detecting a diverse range of threats or prohibited items carried by passengers. This study introduces an advanced deep learning model leveraging YOLOv8, renowned for its enhanced efficiency in automating baggage detection processes. To enhance the model's hyperparameters and adjust them finely during the training process using the baggage dataset, the system utilized a metaheuristic optimization algorithm known as Evolutionary Genetic Algorithm, which is based on evolutionary principles. Incorporating explainable artificial intelligence techniques such as Local Interpretable Model-Agnostic Explanations (LIME) and Gradient-weighted Class Activation Mapping (Grad-CAM) allows for visual interpretation of predictions, aiding operators in utilizing the model effectively. We trained and tested the baggage dataset, which included 8,312 images and five classes: gun, knife, pliers, scissors, and wrench. The YOLOv8 model achieved the following metrics for the detection of prohibited objects in baggage inspection: an overall precision of 90.5%, recall of 83.3%, mAP50 of 91.3%, and mAP50-95 of 67%. The proposed method can fully automate the recognition of prohibited objects during baggage inspection. This approach is beneficial for designing an integrated, automatic, and non-destructive X-ray image-based classification system.]]>
Copyrights © 2026
