Maneuvering Multitargets Tracking System Using Surveillance Multisensors

Multisensors multitargets tracking is an important task, and the difficulty is how to form the real trajectory of each target accurately. In this article, we construct a complete system for tracking highly maneuvering multitargets in the case of multisensors asynchronous sampling. Our system divides the entire tracking task into three modules: space-time calibration, point-trajectory data association, and trajectory data fusion. An improved least-squares (LS) virtual fusion method is proposed to correct the asynchronous sampling time, and the fast vector calibrates the multisensors' spatial state. In the point-trajectory data association module, an adaptive K-nearest neighbors (AK-NN) algorithm is proposed, which employs the adaptive threshold forming multiple trajectories. In trajectory data fusion, a CV+CT+S multimotion model is proposed with the multisensors' probabilistic data association filter (PDAF) algorithm to track highly maneuvering multitargets actively. The results show that our system performs accurately for moving maneuvering multitargets tracking in complex situations.

Automated summary

In this article, a complete system for tracking highly maneuvering multitargets in the case of multisensors asynchronous sampling is constructed, which is divided into three modules: space-time calibration, point-trajectory data association, and trajectory data fusion. An improved least-squares virtual fusion method and adaptive K-nearest neighbors algorithm are used to enhance tracking accuracy, showing accurate performance for moving maneuvering multitargets tracking in complex situations.