Multisensor Measurements Clustering Using Affinity Propagation

Abstract

Motivated by multiple target track initialization in sensor network, this paper presents a multisensor measurements clustering algorithm based on affinity propagation with association constraint. Measurements from the same sensor cannot be grouped into the same cluster, referred as to association constraint, is considered. As a result, the affinity propagation based clustering algorithm is extended to clustering algorithm with association constraint. The max-sum belief propagation is utilized to produce neighborhood maximum clusters. The proposed method can automatically estimate the number of clusters and its complexity scales quadratically in the number of sensors and the number of measurements. The performance of the proposed method is verified by simulated data.