Innovative Sensing Techniques for Urban Operations

Algorithms will be generated to discriminate among humans, land animals and surface vehicles and to characterize dismount activities in advanced radar systems that can detect and track low RCS, slowly moving targets. These algorithms will utilize both micro-level features (e.g. Doppler spectra, short-time Fourier transforms, independent components and principal components) and macro-level features (e.g. movement pattern, speed, location, time of day). The algorithms will be designed to operate in waveform-agile radars and to exploit urban phenomena such as multipath reflections form surfaces such as flat walls. The algorithms will be developed using a mixture of data that: 1) has been identified and is available, 2) is being collected by TSC for DARPA with an ultra-high resolution radar, or 3) will be synthesized using TSC’s commercially available Radar Image Generator software in conjunction with videos that characterize the size and kinesiology of humans and animals. The discrimination algorithms will be developed using Bayesian classifiers and neural networks. The algorithms will be designed for use in radars operating from UHF to Ka-band, and to work within each radar’s parameters (e.g. signal-to-noise ratio, range and Doppler resolutions, target revisit time).