改进的快速自适应分块均值漂移算法

Developed Adaptive Block Mean-shift Algorithm with Two-frame Method

为了实现复杂情况下的目标跟踪,该文提出了改进的自适应分块策略。通过主分量分析的方法确定不同时刻帧的矩形框大小,然后在不同的情况下根据一定的条件,决定不同的分块方法,从而提高目标在旋转、遮挡、缩放等复杂环境下的跟踪成功率。同时,为了减少寻找目标的时间,该文还提出两帧法,即通过前两帧计算出的质心象素点的个数差,快速预测并确定当前帧中目标可能的质心位置。实验表明,这两种方法相结合不仅有效提高了跟踪的成功率,而且减少了迭代次数,提高了算法的跟踪效率。

In order to adapt the cases of tracking which under the complex environment , this paper propose a developed adaptive block strategy which determine a different block strategy according to different conditions . These conditions include the different tracking windows which determined by principal component analysis of different frames.It will develop the success rate of tracking under a complex environment with rotation , occlusion and zooming .At the same time , in order to reduce the tracking time , this paper proposed two-frame method（ TFM） which compute the pixels between the centroid of the previous two frames and quickly forecast the probable centroid position of current frame .The experiment results show that these two methods not only effectively develop the success rate of tracking , but also reduce the iteration times so that reduce the compute capacity .