Performance evaluation for target tracking systems

Different from simulation-based algorithm analysis, real system evaluation needs to determine the tracks' source derivations. A performance evaluation method for real target tracking systems was proposed. Firstly, the statistical distance between tracks and measurements was formulated and the source derivations of the tracks at each scan were further confirmed. Then, the source derivations of the tracks were determined based on maximum likelihood estimation principle. Besides, three indexes were proposed, including Track Initiation Probability, False Track Initiation Probability and Track Lost Probability, and their formulations were given respectively. In the simulation of Over-the Horizon Radar target tracking, IMMPDA algorithm was considered. The result of method gets in low target probability and high false alarm environment is consistent with that of simulation-based algorithm analysis.