在目标模型的建立过程中,将积分直方图引入到粒子滤波跟踪框架中,提出了一种快速的颜色直方图计算方法,极大地提高了粒子滤波跟踪算法的实时性。为了进一步提高算法的鲁棒性,引入了一种基于邻域颜色特征的匹配搜索机制,当跟踪精度下降时,能够对跟踪结果进行优化,减小跟踪误差。实验结果表明了该算法的有效性。

In this paper, a fast calculation method of color histogram based on the effective combination of integral histogram and kernel-color histogram is presented during building the target template. Based on the fast calculation method, the real-time performance is improved greatly. To further improve the algorithm robustness, a matching searching method based on neighborhood color feature is proposed to optimize the tracking result and reduce the tracking error when the filtering accuracy has decreased. Experimental results show that the proposed algorithm is effective.