A modified adaptive track fusion approach

Track-to-track fusion is an important part of multisensor multitarget tracking. Much research has been done in this area. An adaptive approach for track fusion in multisensor environment proposed by C. Beugnon et al. is investigated in this paper. The algorithm chooses the method for calculating the global estimate according to a decision logic, which is based on comparison between distance metric and threshold. Unfortunately, we found that the algorithm, in deriving distance metric, is established under an implicit assumption that sensor level tracks are uncorrelated with global tracks. However, even without process noise the global track and sensor-level track are cross-correlated because they are based on common data. Based on this, a modified adaptive track fusion approach is developed in this paper. The crosscorrelation between sensor-level and global tracks is taken into account in the modified approach. The modified approach still reserves the flexible ability to react to the change of sensor system and it also provides a natural link between track association and fusion. Simulation result illustrates that the modified approach is more robust to the change of system environment.