基于特征融合的Mean-shift算法在目标跟踪中的研究

Study on Target Tracking Based on Characters Inosculation Mean-shift Algorithm

针对仅采用颜色特征的Mean-shift跟踪算法,在目标跟踪过程中,运动目标受光线变化的影响时,易发生跟踪错误等问题,提出一种基于特征融合的Mean-shift算法。新算法结合了颜色特征和边缘方向角特征,并对各特征模型进行加权处理,有效解决了Mean-shift算法在光线突变场景中跟踪失效的问题。实验表明,该算法能够有效提高跟踪的稳健性,达到快速、实时的跟踪效果。

Mean-shift tracking algorithm based on color feature often fails during the course of target tracking, when the target tracked is affected by the change of light. To deal with this problem, a new target tracking approach of Mean-shift is proposed based on characters inosculating of color and edge grads direction. Every character model is dealt with adding a value. Tracking failure in the background with light change can be settled by the new track- ing approach. The experimental results show that the proposed method effectively improves the accuracy and robustness of tracking and achieves the purpose of quickly real time tracking.