DETECTING GUNS, KNIVES, PLIERS, SCISSORS, AND WRENCHES IN X-RAY BAGGAGE IMAGES USING YOLO

Abstract

Aviation security is a top priority in the modern aviation industry. The detection of dangerous items in passenger baggage through Xray imaging currently relies primarily on manual inspection, leading to slow processing speeds and high error rates. This thesis proposes an automated detection system using deep learning to identify five types of dangerous items: guns, knives, pliers, scissors, and wrenches. We employ the YOLOv11 architecture with a FasterNet backbone and propose a targeted augmentation method based on per-class performance analysis. Experimental results demonstrate that the system achieves mAP@0.5 = 90.6% with particularly strong performance in detecting high-threat items (Gun: 98.7%, Knife: 90.4%). This research validates the feasibility of deploying AI systems in airport security screening.