NG-RC Based Multi-Target Data Association Algorithm

The data association problem is the most challenging task in the field of multi-target tracking at sea. Traditional data association methods often fail to balance tracking accuracy and computing speed, while deep learning-based data association methods require a lot of training data. In this paper, we propose a multi-target data association algorithm based on Next Generation Reservoir Computing (NG-RC). Firstly, virtual measurement and distance matrices are introduced to define data correlation models for false alarms and missed detections. Subsequently, the NG-RC network is used to learn the association probability between measurements and existing targets, and the association of multiple targets and multiple measurements is completed in the tracking process. Finally, the simulation results denote that the method can accomplish the multi-target tracking task quickly and accurately in a cluttered environment.