Abstract:

With the emergence of a large number of complex tracking scenarios,conventional multi-source sensor multi-target tracking algorithms have shortcomings of high computational complexity,low tracking accuracy,and inability to estimate target trajectories,making them unable to meet the needs of modern warfare.In this paper,we focus on the implementation of the multi-source sensor tracking problem with the background of the multi-source sensor system composed of active and passive sensors.For the problem of a “multi-active+multi-passive” multi-source sensor system that the measurement cannot be fully integrated and the overall algorithm complexity is high,a multi-source sensor box particle labeled multi-Bernoulli(MS-BPF-LMB)filtering algorithm is proposed.First,the sensors are grouped according to different active sensors,i.e.,all sensors are divided into several "single active + multiple passive" sensor groups;and then,through parallel operations,a multi-sensor information fusion method based on angle correlation is applied to each sensor group to obtain the effective measurements required for tracking.Finally,in the tracking filtering stage,the obtained measurement points are divided into several box particles by introducing the box particle filtering numerical calculation method,and the update coefficients of multi-sensor measurements under box particle filtering are redefined to achieve LMB filtering with a low computational complexity.Simulation results show that the proposed method can effectively deal with the problem of multi-source information fusion of heterogeneous data by significantly reducing the error and decreasing the complexity of the algorithm by about 40% on the basis of maintaining the tracking accuracy of the target.

Key words: target tracking, sensor data fusion, information fusion, box particle filtering, labeled multi-Bernoulli filtering