自适应核相关滤波的运动目标跟踪

Adaptive kernelized correlation filter for object tracking

为实现复杂环境下的目标跟踪,提出了基于自适应核相关滤波(AKCF)的方法。该方法训练状态转移滤波器和尺寸滤波器分别用于目标位置检测和尺寸变化检测。为适应不同运动速度的目标,跟踪过程中利用卡尔曼滤波估计目标运动速度,以改变目标搜索图像块大小。为适应目标尺寸变化,对尺寸滤波器及尺寸因子实时更新。最后将方法与其他相关算法在典型视频序列中进行验证。实验结果表明,所提方法跟踪性能稳定,实时性强,解决了跟踪过程中的目标运行速度变化及尺寸改变等问题。

To realize object tracking in complex environment,adaptive kernelized correlation filter is proposed.The algorithm includes two filters:A translation filter for tracking object and a scale filter for detecting the scale changes.To track the object with different speeds,the searching area is adjusted with respective to the estimated speed which is obtained by using the Kalman filter.To a successfully track the object with different scale,the scale filter and scale factor are updated.Finally,the proposed method was compared with the other state-of-the-art algorithms on several classical videos.The experiment results illustrated that the proposed method performed well especially in case of variations on illumination,pose and scale,different speeds,and partial occlusion.