具有非高斯相关噪声的目标跟踪

Target robust tracking in clutter environment with non-Gaussian related noise

为实现强杂波背景下视频的鲁棒跟踪,在常用非线性系统模型的基础上引入柯西高斯混合噪声模型,充分考虑了非高斯噪声前后时刻的状态相关性,并以权重条件最小方差为标准,推导了非高斯相关噪声的最优建议分布函数,在粒子滤波框架内实现了非高斯相关噪声模型时系统状态的准确估计。在新算法的框架内采用多特征自适应融合的方法,实现了强噪声背景下视频目标的鲁棒跟踪。实验结果表明,本文方法扩展了粒子滤波的适用范围,有效提升了强噪声环境下视频目标跟踪的精度和稳定性。

In order to realize robust tracking of vehicles in the background of the strong clutter, this paper proposes a new kind of particle filter with non-Gaussian and dependent noise. The target vehicle movement behavior in the clutter environment is modeled by introducing Cauchy-Gaussian mixture noise model into the common nonlinear system model. The nature of noise correlation proposal distribution function is analyzed indetail based on the established motion model, and the mathematical expression of non- Gaussian noise correlati6n optimal proposal distribution function is deduced based on the minimum variance weighting conditions. The detailed implementation steps of the algorithm are shown. Finally,robust tracking of moving vehicles in the background of strong noise is realized in the framework of the proposed method based on multiple features adaptive fusion method. Experimental results show that the proposed method extends the application of particle filter, and effectively improves the vehicle tracking accuracy and stability in the strong noise environment.