Consistent Fusion for Distributed Multi-Object Tracking with Different Limited Fields-of-View

The distributed multi-object tracking (DMOT) technologies have attracted much attentions, where multiple distributed sensor nodes sense objects individually and fuse information cooperatively to track multiple objects. However, the multi-node information fusion for DMOT faces the label mismatching problem and different limited fields-of-view (FoVs) problem. To address these problem, in this paper, we propose a distributed multi-object tracking scheme called consistent fusion by associated groups (CFAG) algorithm, including two stages: 1) Distributed track association, which firstly identifies node pairs with overlapping FoVs and then match the tracks by node pairs based on the matching history, and finally the associated track is formed into groups. This stage can reduce the computation complexity when the number of nodes and objects is large. 2) Intra-group fusion, which firstly fuse state estimates based on the arithmetic average within each group, and then label the fused state estimate according to the matching history. For different groups, intra-group fusion can be conducted in parallel. Simulation results indicate that the proposed algorithm performs better than the benchmark in terms of tracking accuracy and label consistency.