Distributed Multi-Object Tracking Under Limited Field of View Sensors

We consider the challenging problem of tracking multiple objects using a distributed network sensors. In practical setting nodes with limited field views (FoVs), computing power and communication resources, we develop novel multi-object algorithm. To accomplish this, first formalise concept label consistency, determine sufficient condition to achieve it \textit{label consensus approach} that reduces inconsistency caused by objects' movements from one node's FoV another. Second, fusion algorithm fuses local state estimates instead densities. This algorithm: i) requires significantly less processing time than density methods; ii) achieves better accuracy considering Optimal Sub-Pattern Assignment (OSPA) errors over several scans rather single scan; iii) is agnostic techniques, only each node provide set estimated tracks. Thus, not necessary assume maintain densities, hence outcomes do modify Numerical experiments demonstrate our proposed solution's real-time computational efficiency compared state-of-the-art solutions in scenarios. also release source code at https://github.com/AdelaideAuto-IDLab/Distributed-limitedFoV-MOT for method foster developments DMOT algorithms.