Greedy Algorithm-Based Track-Before-Detect in Radar Systems

Track-before-detect (TBD) is an effective technique to improve detection and tracking performance for weak targets. Dynamic programming (DP) algorithm is one of the conventional TBD techniques, and has been widely applied to weak target detection and tracking. However, in radar systems, DP-based TBD faces two main challenges: the computational burden and the complex threshold determination. Greedy algorithm (GA) can avoid the above challenges and provide the same or similar performance with DP by forcing some constraints upon the related parameters. In this paper, a novel GA-based TBD (G-TBD) is proposed for weak target detection and tracking in radar systems. The proposed G-TBD contains a two-stage detection architecture. The first detection is implemented with a low threshold to eliminate many noise cells. The second detection is employed to determine whether target exists or not. Detailed analyses, including complexity, parameter settings, and detection performance, are presented. Additionally, the G-TBD is first applied to single frequency network-based MIMO passive radar in this paper. Simulations and real data confirm the effectiveness of the proposed method.