针对红外弱目标跟踪问题, 提出箱粒子标签多伯努利多目标检测与跟踪(Box Particle Labeled Multi-Bernoulli Detection And Tracking, BOX-LMB-DT)算法, 该算法首先通过使用均值滤波对获得的灰度图像进行降噪处理; 其次, 通过将所有像素处依强度大小进行排序, 选出强度较大的区域作为当前时刻的区间量测; 最后利用箱粒子标签多伯努利滤波(Box-Labeled Multi-Bernoulli Filter, Box-LMB)器对目标进行跟踪.仿真结果表明, 所提箱粒子标签多伯努利多目标检测与跟踪算法能够对多目标的航迹和状态进行稳定有效的跟踪, 且在相同条件下, 相较于区间量测下的LMB粒子滤波, 达到相同的跟踪性能时BOX-LMB滤波运算效率提升了22.59%.

For tracking infrared weak targets, a box particle labeled multi-bernoulli multi-target detection and tracking algorithm was proposed. To begin with, the algorithm uses the mean filter to denoise the grayscale image. Then, the region with higher intensity is selected as the interval measurement at current time by sorting the intensity of all the pixels. Finally, the box particle labeled multi-bernoulli filter is applied to tracking. Simulation are presented to demonstrate that the BOX-LMB-DT algorithm has stable and effective performance. In the same conditions, compared with the LMB particle filter under interval measurement, the operation efficiency of the BOX-LMB filtering is improved by 22.59% when the same tracking performance is achieved.