摘要
针对Mean Shift算法核窗宽固定及粒子滤波计算复杂,实际应用跟踪滞后严重问题,提出一种基于均值偏移与粒子滤波融合的目标跟踪算法。利用均值偏移算法的快速收敛性,迭代计算粒子集,保留权值前15%的粒子构成新粒子集,降低系统计算周期,通过重采样得到大权重粒子,更新粒子集,提高目标定位精度。实验结果表明:该算法能够有效解决跟踪运动目标的问题,降低了算法复杂度,提高了计算效率,实验验证过程中未出现跟踪目标丢失的情况。
For the Mean Shift algorithm,the calculation of nuclear window width and particle filter is complex,and the serious problem of tracking lag actually arises. A target tracking algorithm based on the fusion of mean shift and particle filter is proposed. Using the fast convergence of the mean shift algorithm,the particle set is iteratively calculated,and the first 15% of the particles retaining the weight constitutes a new particle set,which reduces the system calculation period,obtains large weight particles by resampling,updates the particle set,and improves the target positioning accuracy. The experimental results show that the proposed algorithm can effectively solve the problem of tracking moving targets,reduce the complexity of the algorithm,improve the computational efficiency,and there is no tracking target loss in the experimental verification process.
引文
[1] Bar-Shalom Y,Chang K C,Bloom H A P.Tracking a Maneuvering Target Using Input Estimation Versus the Interacting Multiple Model Algorithm[J].IEEE Transaction on Aerospace and Electronic System,1989,16(2):296-300.
[2] Munir A,Atherton D P.Adaptive Interacting Multiple Model Algorithm for Tracking a Manoeuvring Target[J].Radar,Sonar and Navigation,IEE Proceedings,1995,142(1):11-17.
[3] Arasaratnam I,Haykin S,Elliott R J.Discrete-Time Nonlinear Filtering Algorithms Using Gauss-Hermite Quadrature[J].Proceedings of the IEEE,2007,95(5):953-977.
[4] Arasaratnam I,Haykin S.Cubature Kalman Filters[J].IEEE Transactions on Automatic Control,2009,54(6):1254-1269.
[5] Fukunaga K,Hostetler L D.The Estimation of the Gradient of a Density Function,with Applications in Pattern Recognition[J].IEEE Transactions on Information Theory,1975,21(1):32-40.
[6] 刘威,赵文杰,李成.时空上下文学习长时目标跟踪[J].光学学报,2016,36(1):179-186.
[7] Collins R T.Mean-Shift Blob Tracking Through Scale Space[C]//Proceedings of IEEE Computer Society Conferenceon Computer Vision and Pattern Recognition.Los Alamitos:IEEE,2003:234-240.
[8] 彭宁嵩,杨杰,刘志,等.Mean-Shift 跟踪算法中核函数窗宽的自动选取[J].软件学报,2005,16(9):1542-1550.
[9] 张恒,李立春,于起峰.尺度方向自适应 Mean Shift 跟踪算法[J].光学精密工程,2008,16(6):167-173.
[10] 黄轩,张军,詹志辉.基于随机惯量权重的快速粒子群优化算法[J].计算机工程与设计,2009,30(3):647-650,663.
[11] 丁晓凤,尚振宏,刘辉,等.基于Mean Shift的多模板目标跟踪算法[J].计算机工程与应用,2017,53(6):141-144.
[12] 邓文浩,唐立才,张南锋,等.基于改进的帧差法和Mean-Shift结合的运动目标自动检测与跟踪[J].现代电子技术,2016,39(4):108-111.
[13] Imath K,Imath L,Barnard M ? V,et al.Mean-Shift and Sparse Sampling-Based SMC-PHD Filtering for Audio Informed Visual Speaker Tracking[J].IEEE Transactions on Multimedia,2016,18(12):2417-2431.
[14] 焦建彬.视觉目标检测与跟踪[M].北京:科学出版社,2016:121-123.
[15] 王倩,曾庆军,张家敏,等.基于自适应渐消无迹粒子滤波的UnscentedFastSLAM算法[J].计算机应用研究,2019,36(5):1-3.