摘要
针对Meanshift目标跟踪算法对强噪声环境敏感的问题,提出了一种结合稳健估计和传统Meanshift的修正Meanshift算法.通过稳健估计修正传统Meanshift算法的核概率密度函数,提升Meanshift算法的鲁棒性.针对信噪比分别为60、30和0 db的仿真数据,将传统Meanshift算法和修正Meanshift算法的跟踪轨迹准确性和精度进行对比.结果表明,修正Meanshift算法能够实现目标准确跟踪,且跟踪位置的相对误差在1%以下.对于实际运动目标视频数据,所提算法也可以实现实时跟踪定位,克服了传统Meanshift算法目标跟踪丢失的问题.
Aiming at the problem that the Meanshift algorithm is sensitive to the strong noise environment,a modified Meanshift algorithm with combining the robust estimation and traditional Meanshift was proposed. The kernel probability density function of traditional Meanshift algorithm was modified through the robust estimation,and the robustness of Meanshift algorithm got improved. Aiming at the simulation data with signal to noise rate( SNR) of 60,30 and 0 db,the accuracy and precision of the tracking trajectory of both traditional and modified Meanshift algorithms were compared. The results indicate that the modified Meanshift algorithm can achieve the accurate tracking of targets,and the relative error of tracking position is below1%. For the real video data of moving targets,the proposed algorithm can also realize the real time tracking and positioning,and can overcome the target tracking missing problem of traditional Meanshift algorithm.
作者
朱闻亚
ZHU Wen-ya(School of Economics and Management, Wuhan University, Wuhan 430072, China School of Mechanical and Electrical Information, Yiwu Industrial & Commercial College, Yiwu 322099, China)
出处
《沈阳工业大学学报》
EI
CAS
北大核心
2017年第2期177-182,共6页
Journal of Shenyang University of Technology
基金
浙江省高等教育教学改革项目(JG2015343)
关键词
稳健估计
均值偏移
目标追踪
噪声
核概率密度函数
中位数
误差
视频
robust estimation
Meanshift
target tracking
noise
kernel probability density function
median
error
video