粒子滤波硬件实现的快速残差再采样策略

High-speed Residual Resampling Scheme for Hardware Implementation of Particle Filter

粒子滤波适用于非线性模型和非高斯噪声的目标跟踪,但其大计算量,限制了实时应用。硬件实现是解决此问题的有效途径。本文针对粒子滤波的第三个步骤-再采样的硬件实现,引入动态基准位置和动态阈值设置,将“粒子-标签”残差再采样的高效定点实现方法进行扩展,使其可以直接对非归一化的权值进行再采样。免去需要进行大量除法运算的权值归一化操作,提高粒子滤波的实时性。实验结果表明其有效性。

particle filter has been proved to be a powerful tool for object tracking in nonlinearity and non-Gaussianity setting,but large computational cost limits its real-time application. Hardware implementation should be an efficient approach to overcome this drawback. For hardware implementation of resampling-the third stage of particle filter, dynamic benchmark position and dynamic threshold are introduced, extending ＂particle-tagging＂ method to process non-normalization weights directly, avoiding a large number of division calculations for normalization, and increasing real-time performance of particle filter. Experiment results demonstrate its validity.