With consideration of low tracking accuracy and even losing target when the track-before-detect method based on particle filter (PF-TBD) tracks infrared dim small target in the complex background. In this article, a track-before-detect method based on the spring model firefly algorithm optimization particle filter (SFA-PF-TBD) was proposed to address this problem. First, the attractiveness and movement behavior of fireflies were introduced into the particle filter for optimizing the particles, and the optimization strength was controlled by evaluating the real-time distribution of particles. After optimization, detecting the density of the particles around the optimal particle, the elastic mechanism of spring was used to control the density of particles around the optimal particle when the particles gathered excessively, which made the distribution of particles more reasonable. Then, the TBD method was realized by the improved particle filter for tracking dim small target under the low signal-noise ratio conditions. Finally, we compared the SFA-PF-TBD algorithm with other algorithms by tracking experiments in simulation scenes and actual scenes. The results showed that the SFA-PF-TBD algorithm has more advantages than the PF-TBD algorithm.

中文翻译：

---

用于红外弱小目标检测与跟踪的智能粒子滤波器


考虑到基于粒子滤波器的先跟踪方法（PF-TBD）在复杂背景下跟踪红外弱小目标时，跟踪精度较低，甚至丢失目标。本文提出了一种基于弹簧模型萤火虫算法优化粒子滤波器（SFA-PF-TBD）的检测前追踪方法来解决这一问题。首先，将萤火虫的吸引力和运动行为引入粒子滤波器对粒子进行优化，通过评估粒子的实时分布来控制优化强度。优化后，检测最优颗粒周围的颗粒密度，当颗粒过度聚集时，利用弹簧的弹性机构来控制最优颗粒周围的颗粒密度，使颗粒分布更加合理。然后，通过改进的粒子滤波器实现了低信噪比条件下跟踪弱小目标的TBD方法。最后，我们通过模拟场景和实际场景的跟踪实验，将SFA-PF-TBD算法与其他算法进行了比较。结果表明，SFA-PF-TBD算法比PF-TBD算法更具优势。