Social demands on public safety are increasing, and the requirements for security measures have become more focused on not impeding the flow of people and not forcing them to a specific posture or motion for inspections. This paper presents the development of a new walk-through security screening system called Invisible Sensing (IVS), which detects dangerous objects hidden in baggage or under clothing in a non-contact manner without making people stop. The paper also describes the key enabling technologies and their implementation and demonstrates the performance of proposed technologies using developed equipment.

The IVS system utilises the penetration characteristics of radio waves to generate images through radar signal processing and uses deep learning to detect weapons or dangerous objects in radar images. The system scans a subject based on the proposed multi-region screening, where the scanning area is optimised by considering the positional relationship between the radar Tx and Rx antennas and a walking human subject. In radar imaging, a motion blur suppression (MBS) technique is proposed, which constructs a high-quality radar image by compensating for the motion blur caused by a person’s walk. In object detection, the proposed approach takes multi-view projected images across multiple regions by compressing the huge amounts of 3D complex data measured by the radar, which enables real-time whole-body screening.

A developed IVS system prototype that integrates and implements these radar measurement, image construction and detection functions is demonstrated. The proposed technologies have been validated in real-time operation with the developed prototype. It has been shown that the proposed imaging with MBS can construct a high-quality radar image of a walking person and achieve good performance in whole-body screening with the proposed multi-view fast detection technology.

This paper deserves high praise as it not only proposes innovative technologies such as motion-robust radar imaging and fast object detection but also presents their practical implementation into an integrated system and demonstrates the world’s top-level throughput (up to 3,600 persons/hour) and empirical detection performance in real-time operation. The fact that the system has shown a path to solving social problems is highly commendable. Future social deployment is expected.