正在加载图片...
·254· 工程科学学报,第37卷,第2期 表2典型帧的平均跟踪误差比较 Jiang M X,Wang H Y,Liu X K.A multi-target tracking algo- Table2 The comparison of average tracking erors of typical frames rithm based on multiple cameras.Acta Autom Sin,2012,38(4): pixels 531 典型帧 100200300400500600700 (姜明新,王洪玉,刘晓凯.基于多相机的多目标跟踪算法 自动化学报,2012,38(4):531) Zhang的方法9.72.10.3000.55.4 B3]Xu H X,Wang Y N,Yuan X F,et al.A hierarchical mean shift Jiang的方法 4.75.210.5113.6 algorithm for object tracking.Acta Autom Sin,2009,35(4):401 本文方法 3.61.2 0 000.21.6 (许海霞,王耀南,袁小芳,等.一种分层Mean Shift目标跟 踪算法.自动化学报,2009,35(4):401) 30 4]Zhou H R.A survey of multiple targets tracking technique.Acta Aeronaut Astronaut Sin,1986,7(1):1 2 一Zang的方法 (周宏仁.多目标跟踪技术综述.航空学报,1986,7(1):1) 一Jiang的方法 20 本文方法 5]Yu Q,Medioni G.Multiple-arget tracking by spatiotemporal Monte Carlo Markov chain data association.IEEE Trans Pattern A- nal Mach Intell,2009,31(12):2196 6]Serratosa F,Alquezar R,Amezquita N.A probabilistic integrated object recognition and tracking framework.Expert Syst Appl, 2012,39(8):7302 400 500 [7]Maggio E,Taj M,Cavallaro A.Efficient multi-arget visual track- 时间帧 ing using random finite sets.IEEE Trans Circuits Syst Video Tech- 图5平均误差对比 nol,2008,18(8):1016 Fig.5 Comparison of average tracking errors 8] Sharp I,Yu K.Sathyan T.Positional accuracy measurement and error modeling for mobile tracking.IEEE Trans Mobile Comput, 4结论 2012,11(6),1021 ] Jiang H,Fels S,Little JJ.A linear programming approach for (1)针对由遮挡、杂波扰动和光照变换带来的跟 multiple object tracking /Proceedings of Conference on Computer 踪不稳定问题,设计了一种新型基于网络流模型的跟 Vision and Pattern Recognition.Minneapolis,2007:744 踪器。模型应用整数规划理论与最小费用流算法相结 [10]Zhang L,Li Y,Nevatia R.Global data association for multi-ob- 合的信息素获取方式,将流费用的变化与目标移动进 ject tracking using network flows /Proceeding of 26th IEEE Conference on Computer Vision and Pattern Recognition.Anchor- 行合理关联 age,2008:342 (2)针对本文所提模型的可行性给出理论支持, [11]Cook W J,Cunningham W H,Pulleyblank W R,et al.Combi- 证明了应用于本文跟踪器的整数规划模型在线性松弛 natorial Optimization.Beijing:Higher Education Press,2011 后可获得原整数假设的全局最优解 (Cook W J,Cunningham W H,Pulleyblank W R,等.组合优 (3)将本文提出的基于最小费用流建模的目标跟 化.北京:高等教有出版社,2011) 02] 踪器应用于PETS09视频库中进行验证,与近期相关 Bugeau A,Perez P.Track and cut:simultaneous tracking and segmentation of multiple objects with graph cuts.EURASIP JIm- 方法相比,在保证跟踪结果的同时,本文方法可大大减 age Video Process,2008:317278 少平均跟踪误差:在同一规划时间尺度下,本文设计的 13] Bertsekas D P.Conrex Optimisation Theory.Beijing:Tsinghua 跟踪器可求得精度稍高的解 University Press,2011 (Bertsekas D P.凸优化理论.北京:清华大学出版社,2011) 参考文献 [14]Maddalena L,Petrosino A.Stopped object detection by learning [Kim I S,Choi HS,Yi K M,et al.Intelligent visual surveillance: foreground model in videos.IEEE Trans Neural Netuorks Learn a survey.Int J Control Autom Syst,2010.8(5)926 Sst,2013,24(5):723工程科学学报,第 37 卷,第 2 期 表 2 典型帧的平均跟踪误差比较 Table 2 The comparison of average tracking errors of typical frames pixels 典型帧 100 200 300 400 500 600 700 Zhang 的方法 9. 7 2. 1 0. 3 0 0 0. 5 5. 4 Jiang 的方法 4. 7 5. 2 1 0. 5 1 1 3. 6 本文方法 3. 6 1. 2 0 0 0 0. 2 1. 6 图 5 平均误差对比 Fig. 5 Comparison of average tracking errors 4 结论 ( 1) 针对由遮挡、杂波扰动和光照变换带来的跟 踪不稳定问题,设计了一种新型基于网络流模型的跟 踪器. 模型应用整数规划理论与最小费用流算法相结 合的信息素获取方式,将流费用的变化与目标移动进 行合理关联. ( 2) 针对本文所提模型的可行性给出理论支持, 证明了应用于本文跟踪器的整数规划模型在线性松弛 后可获得原整数假设的全局最优解. ( 3) 将本文提出的基于最小费用流建模的目标跟 踪器应用于 PETS09 视频库中进行验证,与近期相关 方法相比,在保证跟踪结果的同时,本文方法可大大减 少平均跟踪误差; 在同一规划时间尺度下,本文设计的 跟踪器可求得精度稍高的解. 参 考 文 献 [1] Kim I S,Choi H S,Yi K M,et al. Intelligent visual surveillance: a survey. Int J Control Autom Syst,2010,8( 5) : 926 [2] Jiang M X,Wang H Y,Liu X K. A multi-target tracking algo￾rithm based on multiple cameras. Acta Autom Sin,2012,38( 4) : 531 ( 姜明新,王洪玉,刘晓凯. 基于多相机的多目标跟踪算法. 自动化学报,2012,38( 4) : 531) [3] Xu H X,Wang Y N,Yuan X F,et al. A hierarchical mean shift algorithm for object tracking. Acta Autom Sin,2009,35( 4) : 401 ( 许海霞,王耀南,袁小芳,等. 一种分层 Mean Shift 目标跟 踪算法. 自动化学报,2009,35( 4) : 401) [4] Zhou H R. A survey of multiple targets tracking technique. Acta Aeronaut Astronaut Sin,1986,7( 1) : 1 ( 周宏仁. 多目标跟踪技术综述. 航空学报,1986,7( 1) : 1) [5] Yu Q,Medioni G. Multiple-target tracking by spatiotemporal Monte Carlo Markov chain data association. IEEE Trans Pattern A￾nal Mach Intell,2009,31( 12) : 2196 [6] Serratosa F,Alquezar R,Amezquita N. A probabilistic integrated object recognition and tracking framework. Expert Syst Appl, 2012,39( 8) : 7302 [7] Maggio E,Taj M,Cavallaro A. Efficient multi-target visual track￾ing using random finite sets. IEEE Trans Circuits Syst Video Tech￾nol,2008,18( 8) : 1016 [8] Sharp I,Yu K,Sathyan T. Positional accuracy measurement and error modeling for mobile tracking. IEEE Trans Mobile Comput, 2012,11( 6) ,1021 [9] Jiang H,Fels S,Little J J. A linear programming approach for multiple object tracking / / Proceedings of Conference on Computer Vision and Pattern Recognition. Minneapolis,2007: 744 [10] Zhang L,Li Y,Nevatia R. Global data association for multi-ob￾ject tracking using network flows / / Proceeding of 26th IEEE Conference on Computer Vision and Pattern Recognition. Anchor￾age,2008: 342 [11] Cook W J,Cunningham W H,Pulleyblank W R,et al. Combi￾natorial Optimization. Beijing: Higher Education Press,2011 ( Cook W J,Cunningham W H,Pulleyblank W R,等. 组合优 化. 北京: 高等教育出版社,2011) [12] Bugeau A,Perez P. Track and cut: simultaneous tracking and segmentation of multiple objects with graph cuts. EURASIP J Im￾age Video Process,2008: 317278 [13] Bertsekas D P. Convex Optimization Theory. Beijing: Tsinghua University Press,2011 ( Bertsekas D P. 凸优化理论. 北京: 清华大学出版社,2011) [14] Maddalena L,Petrosino A. Stopped object detection by learning foreground model in videos. IEEE Trans Neural Networks Learn Syst,2013,24( 5) : 723 · 452 ·
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有