Relative discriminant coefficient based multi-cue fusion for robust object tracking

Abstract

In visual tracking, integrating multiple cues will increase the reliability and robustness of the tracking system in situations where no single cue is reliable. In this paper, a novel multi-cue based tracking method is presented under the particle filter framework. Considering both practical distance and Bhattacharyya distance between particles and the target, a parameter called relative discriminant coefficient (RDC) is designed to measure the tracking ability for different features. Multi-cue fusion is carried out in a reweighing manner based on this parameter. Experimental results demonstrate the high robustness and effectiveness of our method in handling appearance changes, cluttered background, illumination changes and occlusions.

References

1. Collins R T Liu Y Leordeanu M Online selection of discriminative tracking featuresIEEE Transactions on Pattern Analysis and MachineIntelligence 2005 27(10)16311643. doi:10.1109/TPAMI.2005.205
2. Han B Davis L Robust observations for objecttrackingIn: Proceedings of the 12th InternationalConference on Image ProcessingWashingtonIEEE Press 2005 2442445
3. Yang C Duraiswami R Davis L Efficient mean-shift tracking via a new similarity measureIn: Proceedings of IEEE Computer Society Conferenceon Computer Vision and Pattern RecognitionWashingtonIEEE Press 2005 1176183
4. Comaniciu D Ramesh V Meer P Real-time tracking of non-rigid objects using mean shiftIn: Proceedings of the 1st Conference on ComputerVision and Pattern RecognitionWashingtonIEEE Press 2000 2142149
5. Jepson A D Fleet D J EI-Maraghi T F Robust online appearance models for visual trackingIEEE Transactions on Pattern Analysis and MachineIntelligence 2003 25(10)12961311. doi:10.1109/TPAMI.2003.1233903
6. Comaniciu D Ramesh V Meer P Kernel-based object trackingIEEE Transactions on Pattern Analysis and Machine Intelligence 2003 25(5)564577. doi:10.1109/TPAMI.2003.1195991
7. Li P Francois C Image cues fusion for objecttracking based on particle filterIn: Francisco JBruce AArticulated Motion and Deformable Objects. Lecture Notes in ComputerScienceBerlinSpringer-Verlag 2004 317999107
8. Wang H Suter D Efficient visual tracking byprobabilistic fusion of multiple cuesIn: Proceedings of 18th International Conference on Pattern RecognitionWashingtonIEEEPress 2006 4892895
9. Triesch J Von der Malsburg C Democratic integration: self-organizedintegration of adaptive cuesNeural Computation 2001 13(9)20492074. doi:10.1162/089976601750399308
10. Spengler M Schiele B Towards robust multi-cue integrationfor visual trackingMachine Vision and Applications 2003 14(1)5058. doi:10.1007/s00138‐002‐0095‐9
11. Shen C Van den Hengel A Dick A Probabilistic multiple cue integration for particle filterbased trackingIn: Proceedings of the 7thDigital Image Computing: Techniques and ApplicationsSydneyThe Universityof Queensland 2003 399408
12. Pérez P Vermaak J Blake A Data fusion for tracking with particlesIn: Proceedings of the IEEE 2004 92(3)495513. doi:10.1109/JPROC.2003.823147
13. Loy G Fletcher L Apostoloff N et al.An adaptive fusion architecture for target trackingIn: Proceedings of the Fifth IEEE International Conferenceon Automatic Face and Gesture RecognitionWashingtonIEEE Press 2002 261266
14. Wu Y Huang T S Robust visual tracking by integratingmultiple cues based on co-inference learningInternational Journal of Computer Vision 2004 58(1)5571. doi:10.1023/B:VISI.0000016147.97880.cd
15. Gordon N Salmond D Smith A A novel approach to nonlinear/non-Gaussian Bayesian stateestimationIn: IEEE Proceedings-F 1993 140(2)107113
16. Nummiaro K Koller-Meier E Gool L V An adaptive color-based particle filterImage and Vision Computing 2003 21(1)99110. doi:10.1016/S0262‐8856(02)00129‐4
17. Isard M Blake A Condensation–conditionaldensity propagation for visual trackingInternational Journal of Computer Vision 1998 29(1)528. doi:10.1023/A:1008078328650
18. Bhattacharyya A Ona measure of divergence between two statistical populations definedby their probability distributionsBulletinof the Calcutta Mathematical Society 1943 3599110
19. Brasnett P Mihaylova L Bull D et al.Sequential Monte Carlo tracking by fusing multiplecues in video sequencesImage and VisionComputing 2007 25(8)12171227. doi:10.1016/j.imavis.2006.07.017
20. Goubet E Katz J Porikli F Pedestrian tracking using thermal infrared imagingMitsubishi Electric Research Laboratories TechnicalReport TR126 2005