基于可行性矩阵遍历的重叠多目标角轨迹关联算法

目标角轨迹关联旨在将隶属于同一目标的不同传感器测量的角轨迹正确匹配在一起，角轨迹关联问题是天基光学监视系统的核心难点问题之一，其性能直接决定了天基光学监视系统的目标跟踪性能。在密集多目标场景下，受光学传感器像元分辨率、光学系统点扩散效应、目标相互遮挡等影响，容易发生目标重叠而不可分辨的现象。现行多目标角轨迹关联方法是基于一个目标产生一条角轨迹的假设前提，在重叠多目标情况下关联性能大大降低。针对此类问题，本文提出了基于可行性矩阵遍历的目标角轨迹关联算法。该算法以倾角差为统计量，遍历各种可能的关联分配，并计算每一个可能的关联分配概率，最后基于贝叶斯准则计算得到最终的关联结果。仿真实验表明，该算法能有效处理目标相互遮挡、目标未分辨等导致的多个目标共用一条角轨迹的情况，具有更好的复杂多目标场景适应性。

An Angle Track Association Algorithm Based on Feasibility Matrix Traversal

? ?Angle track association algorithm aims to correctly associate trajectories observed by different sensors based on principle of identity， it is one of the core and difficult problems of space-based optical surveillance system. Its performance directly determines the target tracking performance of space-based optical surveillance system. In the dense multi-target scene， target overlap occurs frequently affected by sensor resolution， the point diffusion effect of the optical system and target occlusion. Current algorithms achieve good performance in most scenes， but they can not deal with the problem of association between one angle track and multiple angle tracks caused by occlusion and unresolved targets， it is no longer satisfies the one-to-one association assumption， and current algorithms achieve poor performance. Current algorithms satisfy the one-to-one assumption and achieve poor performance in the case of occlusion and unresolved targets. To solve this problem， this paper proposes an algorithm based on traversal of the feasibility matrix. The algorithm takes the hinge angle difference as the statistic and considers all the possible association assignments， and calculates the probability of each possible association event， and the final association result is calculated based on Bayesian criterion. Simulation results show that the algorithm can effectively deal with the situation that multiple targets share an angle track caused by occlusion and unresolved of target， and has better adaptability to complex multi-target scenes.