优化的多模型粒子滤波机动微弱目标检测前跟踪方法

在机动微弱目标的检测和跟踪方面，当前主要研究方法之一是多模型粒子滤波检测前跟踪（MMPF-TBD），该方法以尽可能多的运动模型去匹配目标的机动，符合运动模型精细化研究方向，但存在模型数目与类别较多，模型之间转移计算复杂和有效模型使用效率低等问题。本文从多个运动模型结构上的相似性出发，提出一种优化的多模型粒子滤波检测前跟踪方法，通过粒子机动加速度的变化，在一个模型框架下模拟出类似MMPF TBD中的多种机动模型，简化了算法结构；在该方法实现过程中，采用辅助粒子滤波提高状态估计精度。仿真实验表明该方法相比MMPF-TBD具有更稳定的检测和跟踪性能以及在低信噪比环境中更好的适用性。 

Optimized Multiple Model Particle Filter Track-Before-Detect Algorithm for Maneuvering Weak Target

The multiple model particle filter track-before-detect (MMPF-TBD) is currently one of the mainstream research methods in tracking and detection area for maneuvering weak target, which utilizes as many dynamic models as possible to match the motion of the target. This method accords with the research direction of models’ division; however, the complex transformation and low efficiency of the models always trouble us. Considering the similarity of the models in structure, we proposed an optimized multiple model particle filter track-before-detect algorithm in this paper, which generates many models from one framework by the change of particles’ acceleration. It reduces the complexity of MMPF-TBD and improves the precision of state estimation as we use auxiliary particle filter to implement it. Finally the simulation shows it has a more stable performance in detecting and tracking a maneuvering weak target compared with the traditional MMPF-TBD, in addition, it is more applicable in low signal-to-noise environment.