Pixel-Wise Localization of Concealed Objects on Millimeter-Wave Radar Images Using Deep Learning

Abstract

Automatic detection and localization of anomalies on radar images of personnel taken at the airport security checkpoints is a necessary step of having an end-to-end automatic threat detection algorithm. This article presents two deep learning-based solutions for pixel-wise localization of body-worn anomalies. The trained 2-D and semi-supervised U-Net models can accurately detect and localize foreign objects on all body regions by producing anomaly and body masks for each input radar image.