Long-range Target Identification

Robust long range target recognition for non-cooperative targets in low-observable environments requires recognition algorithms that exploit the physics of feasible long range sensors. Conventional methods of processing signals from non-cooperative moving targets extract very little of the useful information encoded in the signals, resulting in features with poor performance. In contrast, three-dimensional analysis allows extraction of stable, pose-invariant features. This analysis will generate high fidelity target models leading to improved assisted target recognition algorithms and the ability to re-acquire targets after obscuration. The Michigan Technological University Research Institute (MTRI), led by Dr. Mark Stuff, and Integrity Applications Incorporated (IAI), propose to develop, refine, and test signal processing algorithms that enhance exploitation of moving vehicles radar returns, allowing feature extraction not available to current processing. As described below, this work will investigate advanced estimation methods for decomposing signals into distinct scattering centers by combing DP approaches with insights and methods from a fielded Navy tracking technology along with local non-linear optimization processes, advances in motion analysis from new de-interference methods, continuum dictionary methods, automated 3D correspondence analysis for target orientations, and target rigidity analysis. BENEFIT: Since the technology will exploit motions of the sensor and/or motions of the target, it can have application to moving sensors (e.g., airborne) or fixed sensors. Thus the potential market for civilian applications includes many situations in which fixed cameras are now used, for which the extension to 24 hour, all weather effectiveness is needed. This includes large scale (e.g., city-wide) surveillance applications that monitor traffic and track vehicles.