摘要
针对基于核相关滤波的目标跟踪算法,在目标发生遮挡或尺度变化时,跟踪精度和成功率不佳的问题,提出一种改进核相关滤波的目标跟踪算法。算法对跟踪目标的表观模型进行改进,将快速梯度直方图特征与颜色特征自适应融合,提高目标特征表达能力。其次,将相关滤波和粒子滤波融合作为跟踪器,在目标发生尺度变化时,增强跟踪器的鲁棒性。除此之外,引入跟踪状态判断机制,判断目标跟踪状态,在判断目标跟踪失败的时候,能快速对目标进行重定位。在公开数据集上将提出的算法与现有优秀算法进行对比试验,实验结果表明,提出的算法能有效改善目标跟踪性能,跟踪准确,鲁棒性强。
Proposes an optimized target tracking algorithm based on kernel correlation filter to solve the problems of poor tracking accuracy and suc⁃cess rate when the target is occluded or size-changing.Firstly,the algorithm optimizes the active appearance model which is used to track the target,improves the ability of target feature expression through the combination of Histogram of Oriented Gradient and Color adapta⁃tion.Secondly,the correlation filter and particle filter are merged as a tracker to enhance the robustness of the tracker when the target scale changes.In addition,introduces the tracking state determination mechanism to judge the target tracking state.The tracker relocates the tar⁃get immediately when the target tracking fails.Comparing with the excellent algorithms existed on public datasets,the proposed algorithm with strong robustness can effectively improve the performance and accuracy of target tracking.
作者
丁健伟
DING Jian-wei(College of Computer Science,Sichuan University,Chengdu 610065)
出处
《现代计算机》
2019年第36期59-63,共5页
Modern Computer
关键词
目标跟踪
相关滤波
粒子滤波
重定位
Target Track
Correlation Filter
Particle Filter
Relocation