New developments in flexible ID association-based tracking algorithm

For robust data association performance, tracking algorithms available in the literature utilize kinematic as well as non-kinematic information. These algorithms, however, do not provide a systematic way to utilize non-kinematic information to resolve severe and prolonged association ambiguities in the past. In a previous work, we proposed a novel framework in which kinematic and non-kinematic information of potential targets are stored as different entities, respectively denoted as tracks and IDs. The dynamic association between tracks and IDs provides the mechanism for resolving past ambiguities and reporting any remaining ambiguity to the user. However, restrictive assumptions of no false alarms and perfect detection of targets were made in our previous work. The current paper generalizes the algorithm beyond those restrictions. Moreover, algorithms for track and ID initialization and deletion are presented. Simulation results are provided to show the effectiveness of the approach.