通过对跟踪目标颜色特征的分析,采用了基于颜色直方图的粒子滤波跟踪算法对目标进行实时跟踪。该算法首先确定目标跟踪模型并归一化颜色直方图,其次设定粒子初始化参数、计算粒子参数、比较颜色直方图从而确定目标位置,最后通过重采样更新粒子。该算法充分利用颜色直方图作为目标的描述特征,同时兼顾粒子滤波对复杂环境的要求,可以实现非线性和非高斯噪声系统的目标跟踪。仿真效果表明,该算法可以很好地对单目标和多目标进行实时跟踪。
The color characteristics of tracking target in video are analyzed in this paper, and an algorithm of the particle filter tracking in real time based on the color histogram is adopted. Firstly ,the target tracking model is determined and the normalized color histogram is calculated. Secondly,the particle' s initialization parameters are set,particle parameters are calculated, color histograms are compared and target location is determined. Finally,the particle is update by resampling. The color histograms as a target of the features are used, complex conditions are required and target tracking is realized in the nonlinear and non-gaussian noise system. The simulation results show that the single object or multiple targets are tracked in real time.