An efficient coherent multi-frame track-before-detect algorithm in radar systems

Multi-frame track-before-detect (MF-TBD) can achieve superior performance for detection and tracking of low signal-to-noise ratio (SNR) targets by performing, upon the intra-frame coherent integration of slow-time samples within a coherent processing interval, the measurement integration between multiple consecutive frames to further accumulate target energy along physically feasible trajectories. Since the scan to scan interval is usually long, the phases of the multiple samples between frames are generally unpredictable (due to the phase ambiguity and the phase change of complex amplitude of the scatter), thus only noncoherent integration is implemented between frames in existing works. In this paper, the coherent integration, both intra and between frames, is studied and referred to as coherent MF-TBD (C-MF-TBD). Firstly, in the maximum likelihood context, the C-MF-TBD problem is built up and translated into an optimization search in a six-dimensional space. Then, a sequential optimization solution algorithm is proposed to solve the high dimensionally optimization in an efficient but accurate manner. Despite the sub-optimality, simulation results show that the proposed C-MF-TBD method significantly outperforms the traditional MF-TBD algorithm (regardless of phase information between frames).