A concealed ammo detection system for passengers luggage screening

Abstract

Airport passenger screening programs mainly involve checkpoint screening using magnetometers to detect metallic weapons on passengers and X-Ray systems to examine carry-on items. The aviation security system is a human system involving extensive reliance on not only the human's perception but also the human's performance, decision making, and judgment. This is particularly true with X-Ray screening systems as they involve human resource intensive activities in order to detect and resolve potential threat items. In addition to the screener's performance which remains a continuing concern, the screening personnel and for many reasons, have long been regarded as a highly fallible and vulnerable element of the aviation screen. Adversaries seeking to carry out unethical measures by carrying explosive devices, bomb making components, ammos, or handheld weapons through screening checkpoints may attempt to exploit various limitations in not only the human's perception and performance but also in the limitations of the current technology capabilities that may compromise security. Although not all security systems are the same, one security hole in airport X-Ray security systems is that they fail to detect all types of ammos known to the industry. In this research, the authors specifically target AK-47 machinegun bullets, 9mm handgun bullets, and 12 gauge shotgun shells, and propose a system that employs digital image processing techniques in order to detect such type of bullets and shells. The proposed system will also assist the security staff by automatically notifying them for any possibility that such a bullet might be carried in a passenger's carry-on item. For testing purposes, the author built a dataset comprising 150 X-Ray images of passenger's luggage that contain different types of bullets and shotgun shells. Simple image processing techniques that includes image enhancement, conversion, labeling of connected components and geometric distance calculations were applied to help in the detection process.