Multi-target tracking in clustered sensor networks based on labeled multi-Bernoulli filtering

In the multi-target tracking (MTT) problem of sensor networks, inconsistencies in sensor fields of view and target labeling can significantly degrade tracking accuracy. These issues are particularly acute in clustered sensor networks. So, this paper presents an MTT algorithm to address the tracking problem in clustered sensor networks with the aforementioned two issues. Specifically, the algorithm comprises three primary components. In the first part, a multi-sensor measurement hypothesis set is constructed for each cluster head (CH) in the clustered sensor network by utilizing the added measurement labels, multi-sensor measurements, and multi-target prediction information. This set is designed to describe the relationships between the measurement information received by the CH from sensors with inconsistent fields of view. Subsequently, a multi-sensor fusion measurement set is obtained by further integrating the multi-sensor measurement fusion method with the acquired hypothesis set. In the second part, the local MTT process for the CH is proposed by incorporating the obtained multi-sensor fusion measurement set with labeled multi-Bernoulli filter. In the third part, the global MTT process for the clustered sensor network is completed by matching and fusing the tracking results of each CH through a constructed non-feedback fusion mechanism. This mechanism is designed to mitigate the impact of target matching errors in target labeling on tracking accuracy. Experimental simulations demonstrate the effectiveness of the proposed algorithm.