A New Efficient Joint Probabilistic Data Association Method for Multiple Target Tracking System

Input of a Multiple Target Tracking (MTT) system is sensor''s detection data include of true and noise observations in tracking space. In MTT an important stage is data association which assigns true observations to the existed tracks. Multiple targets with heavy noise observations can increase the computational complexity and decrease the data association accuracy. Joint Probabilistic Data Association (JPDA) is a conventional algorithm for solving the association problems in the MTT. The JPDA method generates all the possible hypotheses of observation to track associations and their probabilities. Unfortunately, increase in the noise observations and track coalescence lead to the many generations of association hypotheses, complexity computations, low accuracy, slowing or even stopping the JPDA tracking process. In order to resolve this problem, a new method is proposed to improve the JPDA with eliminating the hypothesis’s generation step. The proposed method calculates the data association probabilities using parsing the hypothesis probability, instead of hypothesis generation. Simulation results show that the proposed method has good performance, accurate and fast for MTT in the presence of heavy noise observations.