Advanced search
1 file | 561.64 KB Add to list

Evaluation of background/foreground segmentation methods for multi-view occupancy maps

Author
Organization
Abstract
An occupancy map provides a top view of a scene and can be used for monitoring the activity of people. We estimate occupancy maps using foreground silhouettes from multiple camera views. The ground occupancies computed from each view are fused in a Dempster-Shafer framework. However, it is not clear which background/foreground segmentation method for deriving the silhouettes is most suited for estimating our occupancy maps. We evaluated three segmentation methods from literature (ViBe, gaussian mixture model, method by Petrovic et al.), and one new segmentation method based on the analysis of edges. Occupancy maps were calculated for the APIDIS dataset, and the obtained maps were evaluated using the players' ground truth positions. We found that all methods perform similar in terms of the accuracy of the estimated maps, except the edges based segmentation method which outperforms all other methods. Future work will include texture based segmentation methods, and will focus on robustness with regard to lighting changes.
Keywords
occupancy map, occupancy monitoring, multi-camera, foreground/background segmentation

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 561.64 KB

Citation

Please use this url to cite or link to this publication:

MLA
Van Hese, Peter, Sebastian Grünwedel, Jorge Niño Castañeda, et al. “Evaluation of Background/foreground Segmentation Methods for Multi-view Occupancy Maps.” Proceedings of the 2nd International Conference on Positioning and Context-awareness (PoCA - 2011). Ghent, Belgium: Ghent University, Department of Telecommunications and information processing, 2011. 37–42. Print.
APA
Van Hese, P., Grünwedel, S., Niño Castañeda, J., Jelača, V., & Philips, W. (2011). Evaluation of background/foreground segmentation methods for multi-view occupancy maps. Proceedings of the 2nd international conference on positioning and context-awareness (PoCA - 2011) (pp. 37–42). Presented at the 2nd International Conference on Positioning and Context-Awareness (PoCA - 2011), Ghent, Belgium: Ghent University, Department of Telecommunications and information processing.
Chicago author-date
Van Hese, Peter, Sebastian Grünwedel, Jorge Niño Castañeda, Vedran Jelača, and Wilfried Philips. 2011. “Evaluation of Background/foreground Segmentation Methods for Multi-view Occupancy Maps.” In Proceedings of the 2nd International Conference on Positioning and Context-awareness (PoCA - 2011), 37–42. Ghent, Belgium: Ghent University, Department of Telecommunications and information processing.
Chicago author-date (all authors)
Van Hese, Peter, Sebastian Grünwedel, Jorge Niño Castañeda, Vedran Jelača, and Wilfried Philips. 2011. “Evaluation of Background/foreground Segmentation Methods for Multi-view Occupancy Maps.” In Proceedings of the 2nd International Conference on Positioning and Context-awareness (PoCA - 2011), 37–42. Ghent, Belgium: Ghent University, Department of Telecommunications and information processing.
Vancouver
1.
Van Hese P, Grünwedel S, Niño Castañeda J, Jelača V, Philips W. Evaluation of background/foreground segmentation methods for multi-view occupancy maps. Proceedings of the 2nd international conference on positioning and context-awareness (PoCA - 2011). Ghent, Belgium: Ghent University, Department of Telecommunications and information processing; 2011. p. 37–42.
IEEE
[1]
P. Van Hese, S. Grünwedel, J. Niño Castañeda, V. Jelača, and W. Philips, “Evaluation of background/foreground segmentation methods for multi-view occupancy maps,” in Proceedings of the 2nd international conference on positioning and context-awareness (PoCA - 2011), Brussels, Belgium, 2011, pp. 37–42.
@inproceedings{1215939,
  abstract     = {An occupancy map provides a top view of a scene and can be used for monitoring the activity of people. We estimate occupancy maps using foreground silhouettes from multiple camera views. The ground occupancies computed from each view are fused in a Dempster-Shafer framework. However, it is not clear which background/foreground segmentation method for deriving the silhouettes is most suited for estimating our occupancy maps. We evaluated three segmentation methods from literature (ViBe, gaussian mixture model, method by Petrovic et al.), and one new segmentation method based on the analysis of edges. Occupancy maps were calculated for the APIDIS dataset, and the obtained maps were evaluated using the players' ground truth positions. We found that all methods perform similar in terms of the accuracy of the estimated maps, except the edges based segmentation method which outperforms all other methods. Future work will include texture based segmentation methods, and will focus on robustness with regard to lighting changes.},
  author       = {Van Hese, Peter and Grünwedel, Sebastian and Niño Castañeda, Jorge and Jelača, Vedran and Philips, Wilfried},
  booktitle    = {Proceedings of the 2nd international conference on positioning and context-awareness (PoCA - 2011)},
  isbn         = {9789490705046},
  keywords     = {occupancy map,occupancy monitoring,multi-camera,foreground/background segmentation},
  language     = {eng},
  location     = {Brussels, Belgium},
  pages        = {37--42},
  publisher    = {Ghent University, Department of Telecommunications and information processing},
  title        = {Evaluation of background/foreground segmentation methods for multi-view occupancy maps},
  year         = {2011},
}