基于车辆跟踪的数据关联算法研究

Research on Data Association Algorithm Based on Vehicle Tracking

对于道路车辆目标的跟踪,由于其高机动性、杂波密集等特点,交互式多模型联合概率数据关联无疑是一种好的跟踪算法。但其固有的缺点是计算量会随着目标个数的增多迅速增加,并且其较少的运动模型难以保证跟踪精度,而增加模型个数会引起模型之间的竞争。因此,提出了一种改进的变结构多模型联合数据关联算法(VSMM-JPDA)。理论分析与对比仿真表明:在非机动、机动转弯和直线机动情况下,该算法对目标的跟踪精度都优于固定结构多模型算法,并且运算量大大减少。

For tracking of road vehicle targets,due to its high maneuverability and clutter-intensive characteristics,interactive multi-model joint probability data association is a good tracking algorithm.However,its inherent disadvantage is that as the number of objects increases,its computational complexity will increase rapidly,and its fewer motion models will not be able to guarantee the tracking accuracy,while increasing the number of models will cause competition among the models.Therefore,an improved Variable Structure Multi-Model Joint Data Association Algorithm(VSMM-JPDA)is put forward.The theoretical analysis and comparison simulation show that the proposed algorithm has better tracking accuracy than the fixed structure and multi-model algorithm under non-motoring,maneuvering and linear maneuvering conditions,and the computational cost is greatly reduced.