Bio-Inspired Event Cameras for Missile Detection

Kitware, in collaboration with the University of Dayton, proposes to investigate the use of bio-inspired imaging sensors for autonomous missile detection. Among the various neuromorphic sensor designs, we propose to incorporate Dynamic Vision Sensor (DVS) based event cameras, which perform power-efficient visual monitoring by emulating the human visual system. Instead of densely sampling each pixel at a fixed temporal frequency, as traditional cameras do, DVS-based cameras asynchronously report events at pixels where irradiance has changed significantly. This reduces power consumption in the sensor and produces a sparser set of image data that, in turn, requires less compute intensive processing. In addition to reduced power consumption, this proposal describes event camera advantages that will allow for accurate missile detection against a wider range of backgrounds and reduce the time to alert. In Phase I, we will carry out system design work and develop prototype detection, tracking, and threat recognition algorithms. We will evaluate these to estimate system performance using simulated data, and work with the DoD to develop a surrogate detection problem that will allow for scalable field testing in Phase II and beyond. Approved for Public Release |21-MDA-10789 (21 Apr 21)