改进的抗遮挡MeanShift目标跟踪算法

Improved anti occlusion Mean Shift tracking algorithm

传统Mean Shift目标跟踪算法通过bin-bin颜色直方图表示目标特征,直方图中往往会混入背景颜色信息,造成跟踪不准确;同时由于Mean Shift算法具有局部最优性,当目标受到严重遮挡丢失后,不能对目标重新定位跟踪。为了解决上述问题,在颜色直方图和抗遮挡能力方面进行了改进。利用交叉bin颜色直方图代替传统的bin-bin颜色直方图表示目标特征,减少背景颜色的干扰,提高Mean Shift算法跟踪精度;当目标受到严重遮挡丢失后,通过一种尺度变化调整机制,在全局范围内搜索目标位置,提高Mean Shift算法抗遮挡能力。实验显示,改进后的算法不仅在背景干扰大时对目标的跟踪精度更高,而且当目标受到严重遮挡丢失后,也能够对目标重新定位跟踪。

Traditional Mean Shift tracking algorithm uses bin-bin color histogram which is often mixed with background color information to express target characteristics, causing inaccurate tracking; meanwhile Mean Shift algorithm has local optimality, which cannot reposition and track the object when the object is lost after severe occlusion. In order to solve the above problems, this paper makes improvements from aspects of color histogram and anti-occlusion capability. To improve the tracking accuracy of Mean Shift algorithm, it uses cross bin color histogram instead of traditional bin-bin color histogram expressing target characteristics, reducing the influence of background color. Then, to improve anti-occlusion capability of Mean Shift algorithm when the object is lost after severe occlusion, it uses a scale change adjustment mechanism,searching object position in global scope. Experiments show that, the improved algorithm not only has better tracking performance in the case of big background noise, but also can reposition and track the object when the object is lost after severe occlusion.