A Novel Track Association Method Based on Overlapping Clustering and Evidence Combination

Under the condition of dense tracks and unknown number of targets, how to achieve track association and improve the accuracy of association is still a topic worth studying in the field of information fusion. The traditional track association methods are basically to make pairwise association judgment on the track information. This process is complicated, and the judgment criteria are greatly influenced by human factors, which will interfere with the association results. Clustering algorithm has high efficiency when dealing with large data sets which can be used to solve the above problem. In this paper, a novel track association method based on overlapping clustering and evidence combination is proposed. The distance threshold is introduced to improve the non-exhaustive, overlapping k - means algorithm, and the algorithm is used to cluster the tracks, which realizes the adaptive change of the number of overlapping tracks. The different clustering results gained by adjusting the distance thresholds are converted into evidence, and the final track association result is obtained by using evidence combination. Some experiments demonstrate the effectiveness of the method proposed in this paper.