MICROWAVE IMAGING FOR WALK-WHILE-SCAN SECURITY SCREENING

Microwave imaging is an important tool for security screening applications. Low-power microwave radiation is used to safely and noninvasively form high resolution images of people to screen for concealed threat objects. This is because microwaves easily pass through clothing and strongly reflect off skin and many materials of interest.

Image resolution improves with large aperture size and bandwidth. Commercial imaging systems realize large apertures two ways; as phased array of antennas, or synthetically by mechanically scanning antennas. In both cases long acquisition times permit only one person to be screened at a time while holding a pose. Security checkpoints employing these systems suffer low screening throughput and present a bottleneck that endangers people. Economically increasing screening throughput requires allowing people to move unimpeded while being imaged.

Computational imaging with a frequency diverse aperture provides a path forward. Frequency diverse apertures are composed of antennas designed to have spatially uncorrelated radiation patterns as a function of frequency. A transceiver drives the antennas with a frequency sweep to rapidly take uncorrelated measurements of a scene. A physical model relating transceiver measurements to scene reflectivity is then numerically solved to form an image. In this way hardware complexity is traded for modeling complexity, leveraging computing technology. The resulting system is inexpensive, modular, flat, and has no moving parts.

An experimental microwave imaging system consisting of a frequency diverse aperture driven by a MIMO transceiver operating from 17.5 GHz to 26.5 GHz is described. The imaging system has 24 Tx antennas, 72 Rx antennas, and samples 101 frequency points giving 174528 possible measurement combinations. The transceiver uses an orthogonal coding strategy to acquire complete sets of measurements at 7 Hz, enabling a walk-while-scan modality. Depth cameras are integrated to inform image reconstruction and analysis. Several acceleration strategies are pursued to reduce image reconstruction times. A comprehensive simulation platform is used to optimize system configuration. Near real-time imaging of multiple people in motion is demonstrated.

Images of people walking present unique challenges for automated threat detection. A deformable stitching model for combining images is developed, and a framework for applying the stitching model is proposed.