New analytical approach to detection threshold of a dynamic programming track-before-detect algorithm

Maintaining the constant false alarm rate (CFAR) is an important issue for the dynamic programming-based track-before-detect (DP-TBD) in low signal-to-noise ratio environment. However, the existing method for analysing the false alarm probability, based on extreme value theory (EVT), leads to the inaccuracy of the obtained detection threshold. In this study, a new analytical approach to compute the false alarm probability of DP-TBD is presented. In the proposed method, the generalised Pareto distribution is utilised to approximate the false alarm probability based on the peaks over threshold model, which can maintain CFAR for DP-TBD method in the low signal-to-noise environment effectively. Simulation results show that this approach provides a more accurate false alarm probability estimation than previous EVT methods.