基于连续最小能量的多目标跟踪被引量：1

Continuous energy minimization based multi-target tracking

下载PDF

导出

摘要文中针对多目标跟踪问题,提出基于连续最小能量的多目标跟踪方法。该算法首先利用观测模型、表观模型、运动模型、互斥模型、轨迹维持模型及轨迹修正模型构建一个目标函数;然后利用梯度下降法对构建的目标函数求解,以得到各时刻跟踪目标的近似最小能量及对应的多目标个数和状态;最后采用基于最小能量的智能探测方法,得到平滑、连续的跟踪轨迹。PETS2009/2010benchmark和TUD-Stadtmitte视频序列库实验结果表明,文中提出的方法能够在存在杂波、虚警和漏检的复杂场景中实现多目标的正确关联,得到稳定、持续的跟踪轨迹。 A method to deal with the issue of multi-target tracking by considering continuous energy minimization is proposed. Firstly,observation,appearance,dynamic,mutual exclusion,trajectory persistence,and trajectory regulation models are integrated into an objective function. Then,the gradient descent method is utilized to solve the constructed objective function to obtain the approximate minimum energy at every moment,and obtain the number and status of multi-targets. Finally,the intelligent extrapolation method based on the continuous energy minimization is utilized to achieve the final smoother and more continuous tracking trajectories. Experimental results on PETS2009 /2010 and TUD-Stadtmitte video database demonstrate that the effectiveness and the efficiency of the proposed MAP-HOR scheme.

作者孙娴朱松豪金栋梁

机构地区南京邮电大学自动化学院

出处《南京邮电大学学报（自然科学版）》北大核心 2016年第1期111-118,共8页 Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基金国家自然科学基金(31200496) 国家博士后基金(2014M550297) 江苏省博士后基金(1302087B)资助项目

关键词连续最小能量梯度下降法探测法误警轨迹 continuous energy minimization gradient descent extrapolation false alarm trajectory

分类号 TP391.412 [自动化与计算机技术—计算机应用技术]

引文网络
相关文献

参考文献10

1ANDRIYENKO A, SCHINDLER K. Multi-target tracking by continuous energy minimization[ C ] //IEEE Conference on Computer Vision and Pattern Recognition. 2011 : 1265 - 1272.
2TUZEL O, PORIKLI F,MEER P. Region covariance:A fast descriptor for detection and classification [ C ] //Europe Conference on Computer Vision. 2006:589 - 600.
3AVIDAN S. Support vector tracking[ J ]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26 (8) :1064 - 1072.
4BOCCIGNONE G, CAMPADELLI P, FERRARI A, et al. Boosted tracking in video [ J ]. Signal Processing Letters, 2010,17(2) :129 - 132.
5KUO C, HUANG C, NEVATIA R. Multi-target tracking by on-line learned discriminative appearance models [ C ] // IEEE Conference on Computer Vision and Pattern Recognition. 2010:685 - 692.
6WANG H, HOU X, LIU C. Boosting incremental semi-supervised discriminate analysis for tracking [ C ] // IEEE Conference on Pattern Recognition. 2010:2748 - 2751.
7LIU Y, ZHENG Y, SHEN X. Applying the multi-category learning to multiple video object extraction [ J ]. Pattern Recognition ,2008,42 (9) :2777 - 2785.
8PERERA A, SRINIVAS C, HOOGS A, et al. Multi-object tracking through simultaneous long occlusions and splitmerge conditions[ C]//IEEE Conference on Computer Vision and Pattern Recognition. 2006:666 - 673.
9GREEN P. Reversible jump Markov chain Monte Carlo computation and Bayesian model determination [ J ]. Biometrika, 1995,82(4) :711 - 732.
10MILAN A, ROTH S, SCHINDLER K. Continuous energy minimization for multi-target tracking [ J ]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2014, 36(1) :58 -72.