Context-awareness in Persistent Tracking
Visual tracking in general faces a fundamental dilemma in practice: tracking has to be computationally efficient but verifying whether the tracker is following the true target tends to be demanding, especially when clutter and/or occlusions are present. Thus, many existing methods tend to be either computationally inefficient when using sophisticated image observation models, or vulnerable to distractions when using simple visual measurements. This mainly threatens long-duration robust tracking.

We proposed a novel and powerful solution for real-world tasks. Rather than focusing only on the target, our approach actually tracks a random field where the target motion is one site in the field and the rest are auxiliary objects that are automatically discovered on the fly by video data mining. Auxiliary objects have three properties at least in a short time interval: persistent co-occurrence with the target, consistent motion correlation with the target, and they are easy to track. The collaborative tracking of this random field leads to efficient computation as well as accurate tracking verification. This new approach has exhibited outstanding performance in many challenging real-world testing cases.



Demo Sequences of a Context-aware Tracker
[These videos are for the purpose of demonstrating the technology only. Any reproduction or propagation of these video is explicitly forbidden and the privacy of the persons in the video should be protected to the full extent.]
 
A dance group
A boy in a pool
A mom with the daughter
Note: In each demo: [top-left] results of on-line video data mining that identifies video context of the target (e.g., the headin this case). [top-right] robust information integration that combines all the predictions from the video context.
A boy in a room
A girl and her dady in woods
Note: [bottom-left] comparison with a dedicated head tracker. [bottom-right] our result where the yellow bounding box highlights the output the tracked target.
The major subject of each sequence is a small kid, and the tracking of the head of the kid is of great interest. These sequences exhibist many challenges such as occlusion, background clutter, appearance changes, etc. In addition, these video are recorded by amateurs with a hand-held camera. The tracking task here is very difficult. We developed a method that introduces on-line data mining to visual tracking, which automatically discovers the visual context of the target and the use of the context make the tracking quite stable and robust (for details please see our papers).



Publication:
  1. Ming Yang, Gang Hua and Ying Wu, "Context-Aware Visual Tracking", IEEE Trans. on Pattern Analysis and Machine Intelligence, 2008

  2. Ming Yang, Ying Wu and Shihong Lao, "Intelligent Collaborative Tracking by Mining Auxiliary Objects", in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR 06), New York City, NY, June 17-22, 2006. [PDF]

  3. Ming Yang, Ying Wu and Shihong Lao, "Mining Auxiliary Objects for Tracking by Multibody Grouping", in Proc. IEEE Int'l Conf. on Image Processing (ICIP'07), San Antonio, TX, Sept. 2007 
Return to Tracking Research
Updated 7/2006. Copyright © 2001-2006 Ying Wu