融合目标形状信息及图割窄带优化的目标跟踪算法

提出基于图割窄带优化算法及融合目标形状信息的目标跟踪方法,首先采用卡尔曼滤波方法对目标新的位置进行预测,进而基于目标当前位置及分割结果估计目标的形状的信息,然后在目标预测位置采用窄带的图割优化算法并集成目标的形状先验信息对目标的进行分割,从而确定目标新的位置并得到目标新的轮廓结果,完成目标的精确跟踪。实验结果表明提出的方法具有良好的性能,能够精确有效的跟踪复杂背景中的运动目标。由于采用窄带图割分割优化,使得算法也具有良好的实时性,能够在实际中得到应用。

Object Tracking by Fusing Narrow-Band Graph Cuts and Shape Information

In this paper an object tracking algorithm based on narrow-band graph cuts and object shape information is proposed. Kalman filter is first used to predict the new location of the tracked object, and then the object shape information is estimated based the current object shape. Lastly the narrow band graph cuts is exploited to segment the predicted object and extract the accurate object shape by integrating shape prior into graph cuts in order to track object accurately. The experiments on the real videos demonstrate the good performance of the proposed tracking algorithm. Owing to the narrow band graph cuts the proposed tracking algorithm had good real-time and can be used in practice.