Abstract: An improved particle filter algorithm for object tracking is presented. In order to improve the tracking accuracy and parallelize the computing step, this paper divides the random module into two parts and sorts the weight sequence so that it can choose the threshold adaptively. Additionally, the new architecture proposed takes adventage of the pipelining operation to accelerate the progress of computing and sorting. The results show that the proposed algorithm improves the tracking accuracy and efficiency in both lab scene and occlusion scene.

Key words: particle filter, object tracking, Field Programmable Gate Array（FPGA）

摘要： 提出了一种改进的粒子滤波目标跟踪算法，提出了限定区域的伪随机算法和根据权值分布的自适应重采样算法来提升目标跟踪的精度和并行特性。同时在算法的FPGA硬件结构实现上，对程序结构进行调整，充分利用流水线并行处理数值计算，运用硬件并行特性加快粒子的权值排序过程。实验结果表明，提出的算法在实验室场景与遮挡情况下都具有良好的跟踪准确性和实时性。

关键词: 粒子滤波, 目标跟踪, 现场可编程门阵列（FPGA）