基于K最短路径算法的轨迹关联应用研究

Application of trajectory association based on K-shortest paths algorithm

多目标跟踪可以通过轨迹关联来实现,这种方法具有很强的鲁棒性。然而,当处理多目标跟踪问题时,连接检测目标可能的轨迹簇是一个很难解决的优化问题。使用抽样或贪婪的动态规划方法不能保证得到全局最优解。利用此问题的特殊结构,可以把它重新规划为一个受约束的网络流模型的优化问题,并利用K最短结点不相交路径算法高效地解决,同时得到全局最优解。

Multi-target tracking can be realized by trajectory association. This method has strong robustness. However, when dealing with a multi-target tracking problem, i t is a d if f ic u l t optimization problem to l in k the possible families o f trajectories. The use of sampling-based o r greedy dynamic programming method can’ t guarantee the global optimum. Making use o f the specific structure of this problem, i t can be redesigned as an optimization problem w ith constrained network flow model, and f ind a global optima eff ic iently using the K shortest node-disjoint paths algorithm.