融合视频亮度优化的目标相关滤波跟踪方法

Target correlation filter tracking method for fusion of video brightness optimization

针对相关滤波器在目标跟踪过程中因亮度变化导致输出失效的问题,提出了一种基于视频亮度优化的相关滤波跟踪方法。该方法采用变异及种群多样化策略,提升了粒子群算法的全局寻优能力。通过建立视频序列平均亮度与相关滤波器输出结果方差之间的映射关系,获得不同亮度条件下相关滤波器输出方差的最佳表达;同时,通过计算相邻搜索框之间的相似度来确定相关滤波器模板的更新速度,减少了跟踪失效的概率和计算量。实验证明,优化后的跟踪框架在不同亮度和背景条件下具有更强的目标适应性和鲁棒性。

Aiming at the problem that the output of correlation filter is invalid due to the change of brightness in target tracking process,this paper proposed a correlation filter tracking method based on video brightness optimization. This method used mutation and population diversification strategy to enhance the global optimization ability of particle swarm optimization algorithm. By establishing the mapping relationship between the average brightness of the video sequence and the variance of the output result of the correlation filter,it obtained the best expression of the output variance of the correlation filter under different brightness conditions. At the same time,it determined the update speed of the correlation filter template by calculating the similarity between adjacent search frames,which reduced the probability of tracking failure and the amount of calculation. Experiments show that the optimized tracking framework has better target adaptability and robustness under different brightness and background conditions.