基于模板匹配及区域信息融合提取的快速目标跟踪算法

提出了一种基于融合的快速目标跟踪算法。该方法将目标预测模型、目标模板匹配以及目标空间信息融合到统一框架内。该方法通过预测模型,预测下一帧中目标候选区域,从而降低模板匹配方法的搜索区域。然后在预测模型预测的搜索范围内进行目标的模板匹配,同时,提取目标的区域信息来辅助目标的模板跟踪方法。最终跟踪结果由目标模板跟踪以及目标的空间信息共同决定。本方法由于采用了目标的预测模型,从而能够减少目标的搜索范围,降低因目标的全图搜索造成的时间开销。并且由于减少了搜索区域,剔除了部分相似的匹配窗口,从而提高了跟踪时匹配的精度。本算法结合了目标模板跟踪和目标区域信息的优点,能够增强目标跟踪过程准确度。由于加入了目标的区域信息,对目标跟踪中出现的漂移现象更加鲁棒。实验结果表明,本方法在进行目标跟踪过程中对出现的漂移、遮挡等问题更加鲁棒、稳定。

A fast object tracking algorithm based on template matching and region information fusion extraction

In this paper, we propose a fast object tracking algorithm based on template matching and region information fusion extraction. The method reduces the search candidates based on the prediction framework. Besides, the region information of the object is extracted to assist template based object tracking. The prediction model also reduces the computational complexity of object tracking tasks while the object region information improves the accuracy of object tracking tasks. The proposal combines the advantages of the object template and the object information extraction, thus improving the accuracy of object tracking. The tracking shift and object occlusion are handled using the object motion information and object region information. Experimental results show that the proposed algorithm has robust performance when tracking drifting and blocked objects.