Artificial Intelligence Based Baggage Scanner System

Abstract

The primary objective of an artifcial intelligence-based baggage scanning system is to enhance safety and security in complex environments by effectively detecting contraband items, a crucial measure for public safety. The traditional process for detecting these items is inherently challenging, time-consuming, and prone to oversight. To address these issues, this thesis proposes the adoption of automated deep learning solutions, which represent a signifcant advancement in technology and sophistication. Deep learning, a branch of artifcial intelligence, involves training models on large datasets to recognize patterns from experience and make decisions. Current baggage screening methods are inadequate, prompting the need for automated deep learning solutions. X-ray imaging technology is central to this approach, providing clear images of luggage contents and enabling the detection of hidden dangerous items. Our dataset comprises fve types of potentially dangerous objects: guns, knives, pliers, scissors, and wrenches. This dataset is used to train and test object detection and instance segmentation models. The data is divided into 70% for training, 20% for validation, and 10% for testing. Additionally, data augmentation techniques were employed to enhance the training set, improving the model’s ability to generalize to unfamiliar data instances. The YOLOv8 model was selected for this application due to its architectural improvements over YOLOv5, another version of the YOLO (You Only Look Once) object detection model. YOLO models are renowned for their ability to detect objects quickly and accurately in images, making them ideal for applications like baggage screening. YOLOv8 surpasses YOLOv5 in both accuracy and speed, owing to signifcant architectural enhancements. During extensive testing, the YOLOv8 model, trained on our curated dataset, achieved an impressive 94.8% accuracy in detecting prohibited items. Additionally, YOLOv8 excelled in instance segmentation, with a high mean Average Precision (mAP) of 87.4% in segmenting contraband items. These results underscore the effectiveness and reliability of the selected deep learning models for the specifc application of contraband detection in baggage screening through X-ray imaging.The project shows that using the YOLOv8 deep learning model in baggage scans improves contraband detection with high accuracy and effciency. The use of sophisticated AI and X-ray imaging enhances security measures in high-risk areas by making baggage screening more reliable and effcient.