Suicide Vest Detection Using Deep Learning

Abstract

Current security systems such as walk through gates have an inherent loophole that they provide us with very little reaction time in case of a threat. We propose a standoff arrangement where a radar is projected at a distance (~18m) and it detects hidden objects. The standoff distance gives security personnel necessary reaction time to neutralize the threat. A tri-static configuration is used which emits micrometer waves. The emitted waves when reflected back from a subject have a varied physical orientation. The variation in physical orientation is analyzed and compared with a threshold. We use a deep learning approach in setting up a threshold for declaring the subject as threat or no-threat. Deep learning techniques help us omit background noise and thus eliminating the possibility of false positives.