基于改进的Camshift运动目标跟踪算法的研究

Moving object tracking based on improved Camshift algorithm

针对基于颜色概率分布的连续自适应均值漂移算法(Camshift)跟踪算法在背景中出现相同颜色干扰时容易致使跟踪目标失败的问题,提出了一种改进的Camshift跟踪算法。首先对Camshift跟踪目标前进行目标检测,通过帧差法、光流法、背景差分法三种检测算法对比,采用背景差分法得到的运动目标区域矩形特征参数作为Camshift的初始化参数,取代一般Camshift算法利用颜色特征的跟踪。最后对改进的算法和一般Camshift进行仿真对比实验。实验结果表明,结合背景差分法和连续Camshift算法的运动目标跟踪在一定程度上满足了实时性与稳定性的要求。

Camshift tracking algorithm based on probability distribution of color were susceptible to be interfered by the same color in the background , that lead to the failure of tracking the target. To address this problem, it presented an improved Camshift tracking algorithm. Firstly detecting target before Camshift algorithm tracking, comparising three detection algorithms, the frame difference method, optical flow method, the background difference method getting the rectangular characteristic parameters of moving target area using background difference method as the Camshift initialization parameters, that had replaced the general Camshift tracking algorithm using color features. Finally, it makes comparative experiments between the improved algorithms and general Camshift. The experimental results show that, combined with the background difference method and the continuous Camshift algorithm for tracking of moving objects, to some extent it can meet the real-time and stability requirements.