Appearance Management and Cue Fusion for 3D Model-Based Tracking
   
Jump to: Complete Reference, Abstract, Related Download, BibTeX Entry, Links and Notes, Images, Additional Material
 
Complete Reference
  N. Krahnstoever, R. Sharma, "Appearance Management and Cue Fusion for 3D Model-Based Tracking," IEEE Conference on Computer Vision and Pattern Recognition, Madison, Wisconsin, June 16-22, 2003, accepted for publication.
 
Abstract
  This paper presents a systematic approach to acquiring model appearance information online and to use the acquired information to drive a set of complementary imaging cues to obtain a discriminatory, drift-free, observation model. Appearance is modeled as a Markov Random Field of color distributions over the model surface. The online appearance acquisition process estimates appearance based on uncertain image measurements and is designed to greatly reduce the chance of mapping non-object image data onto the model. Confidences about the different appearance driven imaging cues are estimated in order adaptively balance the contributions of the different cues which allows to maintain performance in the presence of degradation in imaging conditions. The discriminatory power of the resulting model is good enough to allow long-duration single-hypothesis model based tracking under flexible imaging conditions with no prior appearance information.
The performance of the resulting generative model-based tracker is evaluated carefully based on real and semi-synthetic video sequences, showing that the presented algorithm is able to robustly track a vide variety of targets under challenging imaging conditions.
 
Related Download
 

[ PDF (3.3MB) ]

 
BibTeX Entry
 
@inproceedings{krahnstoever03Aappearance,
author={N. Krahnstoever and R. Sharma},
booktitle={Proc. of IEEE Conference on Computer Vision and Pattern Recognition, Madison, Wisconsin, USA},
month={June},
title={Appearance Management and Cue Fusion for {3D} Model-Based Tracking},
year={2003}
}
 
Links and Notes
  CVPR 2003 Conference
 
Images
 
 
Additional Material
  There is also an updated and more detailed Technical Report: [ Hires PDF (11MB) | Lores PDF (6.6MB) ]