一种基于算法融合的运动目标跟踪算法

摘要

To improve the robustness of object tracking algorithm in complex environment, a tracking algorithm based on algorithm fusion was proposed. Firstly, the color and SURF( Speeded Up Robust Features) features were used to presented the interested target. Mean shift algorithm based on color feature was used to find the suboptimal position rapidly. Because the color histogram contains less information, tracking error is accumulated. Then, the SURF feature matching algorithm is used to estimate the center point and size of the target The cumulative errors are amended in time. Finally, the best result is choosed according to the Bhattacharyya coefficient Experimental results show that the proposed algorithm has a good robustness in object deformation, size changing and similar apparent existing around the target and it's real-time.%为了提高目标跟踪算法在复杂环境下的稳健性,提出了一种将基于颜色特征的均值漂移算法和SURF( Speeded Up Robust Features)特征匹配算法相融合的目标跟踪方法.该算法首先采用颜色特征和SURF特征分别描述目标模板,利用均值漂移算法快速估计目标局部最优解.但仅采用单一颜色特征来估计目标位置,跟踪误差逐渐累积；采用SURF算法精确估算目标位置和尺度,及时修正累积误差.最后根据相似性度量Bhattacharyya系数选择较优的结果作为当前帧跟踪结果,且更新目标模板.实验结果表明,算法在目标发生较大形变、尺度变化、周边具有表现相似目标时具有很强的稳健性,且满足跟踪实时性要求.