基于帧间双向互关联的点状多目标检测算法

Probabilistic weighted and bidirectional inter-frame correlation based estimation method for multi-target trajectory initialization

针对图像序列中点目标个数的不确定性和目标航迹的初始化问题,研究了一种利用概率加权和多帧图像间的目标双向互关联特性的先跟踪后检测技术算法:首先对组合帧图像进行基于阀值的首次检测,再对目标群概率加权得出目标的初始位置,利用帧间目标的双向互关联特性对各自目标的航迹初始化。仿真实验结果表明,该检测算法计算量小,实时性好。

For solving the problem of the uncertainty about the targets number in the image sequence and initialize the target tracks,this paper researches on the algorithm Track-Before-Detection（TBD） by using a probability-weighted and multi-frame bidirectional correlation method.Firstly,this paper detects the combined frame based the threshold,secondly,through the association of probability of target groups,the authors can gain. the initial target location,then by using the two-way correlation of interframe goals this paper initializes the track of respective goals.The simulation of results show that the algorithm has a good performance, a small calculation and a good real-time.