A Track Association Algorithm Based on Leader-Follower On-line Clustering in Dense Target Environments

Abstract

The imbalance between accuracy and computational cost is a defect in track association. In response to the defect, the track association problem is transformed into an on-line clustering problem with constraints, and a novel track association algorithm is proposed based on Leader-Follower online clustering. In the algorithm, we take a track as a Leader or a Follower based on its type and make Followers and Leaders clustered, which greatly reduces the track pairs associated. In addition, the association relationships between Leaders and Followers are acquired by introducing a function of association degree, which is characterized by small computational cost and no requirements on the distribution of sensor data. The fused Leader-Follower forms a new Leader, which combines Leader generation and track fusion. When sensor tracks are updated, their Leaders will be changed and the other Leaders will be retained, by which the associated results obtain a good stability.