OMP-Based Target Detection for Pulse-Doppler Radar

Abstract

Radar systems transmit radio-frequency electromagnetic waves and receive the reflected signals from objects within their field of view. A surveillance radar aims to detect and estimate the state of targets in the surveillance volume. Traditional detection algorithms struggle in environments containing closely spaced objects or heavy clutter, resulting in false alarms or missed detections. Greedy compressed sensing algorithms, such as Orthogonal Matching Pursuit (OMP), address this through an iterative process that identifies and subtracts the strongest component in the residual signal, enabling detection of weaker targets masked by stronger ones. In this Master’s thesis, OMP was implemented and evaluated on both real and simulated pulse-Doppler radar data, focusing on scenarios involving closely spaced targets and clutter-rich environments. The performance of the OMP detector was compared to a conventional Cell-Averaging Constant False Alarm Rate (CA-CFAR) detector. A proposed hybrid of the two detectors, denoted CA-OMP, was also evaluated. The results indicate that the OMP detector can resolve closely spaced targets and targets in the presence of strong clutter more effectively than CA-CFAR, particularly in simulated data. However, on real data it is unable to fully suppress detected signal components, resulting in false alarms and degraded detection performance across the surveillance space. The CA-OMP detector exhibited improved detection performance on real data compared to the CA-CFAR detector, while mitigating most of the incorrect detections generated by the OMP algorithm.