检测前跟踪中的目标交叉航迹标签分层管理
Target cross-track labeled hierarchical management for tracking before detection
-
摘要: 针对检测前跟踪算法无法提供目标航迹标签以及传统算法在轨迹交叉情况下批次编号管理混乱的问题, 提出了一种标记关联的航迹管理方法。该方法为各个目标的方位航迹标记不重复的标签, 并针对标签多伯努利算法非共轭先验的问题, 采用将标签多伯努利分布逼近
\delta -广义标签多伯努利分布的方法进行解决。另外还提出一种辅助门限进行分层逐级更新的方法, 实现了对不同的航迹赋予不同的标签, 解决了方位交叉情况下轨迹中断、目标跟丢和错跟的问题。仿真及海试试验表明, 该方法的平均误差稳定且接近为0, 定位精度高且轨迹连续清晰, 不仅能在目标轨迹交叉时进行准确跟踪, 还能实现对多目标数的精准估计, 为基于声呐阵的水下目标态势感知方法提供了新的技术支持。Abstract: In order to solve the problem that the target track label cannot be provided by the track-before-detect algorithm and the batch number management is confused by the traditional algorithm in the case of trajectory intersection, a track management method based on label association is proposed. This method marks the non-overlapping labels for each target’s azimuth track, and aims at the problem of the non-conjugate prior of the labeled multi-Bernoulli algorithm. The method of approximating label multi-Bernoulli distribution to\delta -generalized labeled multi-Bernoulli distribution is used to solve this problem. In addition, an auxiliary threshold hierarchical updating method is proposed to assign different labels to different tracks, which solves the problem of track interruption, target loss and wrong tracking in the case of azimuth intersection. The simulation and sea test results show that the average error is stable and close to 0. The positioning accuracy is high and the trajectory is continuous and clear. Moreover, it can also achieve accurate estimation of the target number. It provides a new technical support for underwater target situation awareness method based on sonar array.