Abstract: The Standard Particle Filter (SPF) is a typical method of solving the tracking problem of non-linear/non-Gaussian model system. However, updating process strictly depends on parameters selection, and it cannot handle the changes in curve topology. In regard to this, a new particle filter target tracking algorithm based on geometric active contours was proposed, which made a good deal with the changes of curve topology using level set theory. The algorithm improved the resampling techniques and increased the diversity of particles. The simulation results indicate that the proposed method can effectively improve the state estimation precision with more flexibility.

Key words: geometric active contour model, Particle Filter (PF), target tracking, resampling

摘要： 标准粒子滤波(SPF)是解决非线性、非高斯模型系统跟踪问题的典型方法,然而粒子更新过程严格依赖于参数的选取,且不能处理曲线拓扑结构的变化。鉴于此,提出基于几何主动轮廓模型的粒子滤波(PF)算法。利用水平集技术处理轮廓曲线拓扑结构变化,改进重采样技术,增加粒子多样性。实验结果表明,该算法是有效可行的,并提高了非线性系统状态的估计精度,具有更强的适应性。

关键词: 几何主动轮廓模型, 粒子滤波, 目标跟踪, 重采样