Radar Target Classification Receiver Using Sparse Regression and Target Tailored Matched Filters

Abstract

Waveform design is a commonly used approach to enhance target classification in high resolution radar systems. In the monostatic case of detecting some extended target, inherent to many of these system models is an assumption of a known target impulse or frequency response. A practical issue with this method is that for nonsimple target cases, such a response can change drastically for varying aspect or viewing angles. In this paper, we first develop a sparse regression method for classifying target class and aspect angle. Next, we derive matched filters tailored to dictionaries of target response profiles representing these variations in aspect angle. The desirable goal of real-time classification is met by suitably reducing the computational cost and not relying on a series of measurement and adaption cycles to achieve classification. The result is a series of filters matched to the target dictionary data and can be used in a hypothesis testing approach to classification. Results are shown for radar cross-sectional data generated from computer-aided design models of different targets.