【摘要】 基于相关滤波的目标跟踪算法通常只利用目标的自身特征,未能充分利用目标周围的背景特征,容易将目标特征误判为背景,从而导致漂移现象的发生.提出一个自适应搜索窗口的相关滤波模型(RIACF),自动调整搜索窗口并有效地利用目标周围的背景信息,显著减少误判情况的发生.为验证算法的有效性,与传统相关滤波算法KCF,CSK及其他优秀的目标跟踪算法Staple,DSST,Struck,TLD,CT进行对比.实验表明:该模型可以显著地提高跟踪器的跟踪效果,虽然该模型引入的背景信息对算法的运行速度有所影响,但该算法仍能保证足够快的运行速度,不影响目标跟踪的实时性.更多还原

【Abstract】 The object tracking algorithms based on correlation filter always extracted the features of object itself but nothing of the background around the object,which resulted in drift phenomenon because algorithms usually misjudge the target features as the background. A new model of adaptive search window based on correlation filter,which could solve the proposed problem and improve the robustness of object tracking was peoposed. In order to verify the effectiveness of the algorithm,it was compared with the traditional correlation filtering algorithms CSK, KCF and other excellent algorithms Staple, DSST, Struck, TLD, CT. The experimental results demonstrated that the proposed tracker could remarkably improve the precision on the challenging benchmark. Although the speed of the algorithm was affected by introducing background information,it still run fast enough to keep in real time.更多还原