一种改进的粒子滤波和Mean Shift联合跟踪算法

为了提高视频运动目标跟踪的准确性和实时性,提出一种改进的粒子滤波和Mean Shift联合跟踪算法.针对传统粒子滤波跟踪算法中颜色直方图观测模型存在的局限性,提出了一种基于分块颜色直方图的观测模型描述方法,并根据该分块直方图的特点,重新设计了粒子权值的更新策略;针对粒子滤波算法实时性差的问题,提出了一种基于积分直方图的颜色特征快速计算方法,极大地降低了算法的运算量;为了降低相似背景干扰对跟踪效果的影响,提出了一种基于Gabor幅度谱的Mean Shift跟踪算法,并利用改进的Mean Shift算法对粒子滤波跟踪结果进行优化,提高了跟踪算法在复杂背景下的搜索能力.实验结果表明了算法的有效性.

An Improved Tracking Algorithm Based on Particle Filter and Mean Shift

To improve the accuracy of video target tracking, a novel joint tracking algorithm using particle filter and mean shift is proposed. Aiming at the shortcomings of color histogram-based observation model in traditional particle filter, a sub-block histogram-based observation model is built. Also based on the new model, weight calculation in particle filter is redesigned. To overcome the bad real-time performance of traditional particle filter, a fast color feature extraction method based on integral histogram is proposed. Furthermore, a Gabor-wavelet-based Mean Shift algorithm which can reduce the interference of similar background is presented, and the improved Mean Shift method is used to optimize the estimated results obtained by particle filtering. Experimental results show that the proposed algorithm is of effectiveness.