An association-based multi-target tracking method

Authors:
Ai Min Li;Pil Seong Park
Affiliations:
Qilu University of Technology Changqing, Jinan, China;University of Suwon Hwaseong-si, Gyeonggi-do, Korea
Venue:
Proceedings of the 2013 Research in Adaptive and Convergent Systems
Year:
2013

Citing 7
Cited 0

CONDENSATION—Conditional Density Propagation forVisual Tracking

International Journal of Computer Vision
Color-Based Probabilistic Tracking

ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part I
Maintaining Multi-Modality through Mixture Tracking

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Detection and Tracking of Multiple, Partially Occluded Humans by Bayesian Combination of Edgelet based Part Detectors

International Journal of Computer Vision
Robust visual tracking for multiple targets

ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part IV
Multi-camera tracking and segmentation of occluded people on ground plane using search-guided particle filtering

ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part III
An adaptive particle filter tracking method based on homography and common FOV

Proceedings of the 2012 ACM Research in Applied Computation Symposium

Quantified Score

Hi-index	0.00

Visualization

Abstract

Multi-target tracking is widely studied, but it is still a difficult problem due to occlusion or split/merge of target images in reality. Solving such problems is the key in multi-target tracking. In this paper, we propose a modified particle filter method to track multiple targets, based on improved target detection and data association. While applying a common particle filter method as the main tracker, another tracker tries to guide and recover it when it fails after long or severe occlusion. This 2nd tracker tries to find candidate targets as accurately as possible by using background subtraction and an AdaBoost classifier detector. Then an ID is assigned to each candidate target by association using the Hungarian method. At last a small number of particles of the main tracker are sampled at the position suggested by the 2nd tracker, which lead the other particles of the main tracker to the correct position quickly. Experiments show a good performance of our method in solving multi-target occlusion problems, and it is proven to be fast enough for real time applications.