Explosives Detection using Shadow Features in Radar Images for Walk-Through Security Screening

Abstract

Radar imaging technologies have been utilized to detect concealed hazardous materials for security screening in public facilities. We have developed a high-throughput walk-through and whole-body security screening system called Invisible Sensing (IVS) based on radar imaging and deep learning. In our previous work, we have demonstrated that the IVS system can detect guns and knives while subject persons walk through the system. This paper presents a newly developed function to detect explosives in radar images on the IVS system. Since most explosives have low reflectivity to microwaves, it is difficult to detect the shape of explosives in radar images. In contrast, the human body is highly reflective and visible in radar images. We propose a novel approach to detect low-reflective explosives in radar images by learning shadow features against the high-reflective human body background. We demonstrate that the proposed detection technique integrated into the IVS system achieved successful explosive detection performance in real time.