Finding the Pose of an Object of Revolution

Abstract

An algorithm able to locate an object of revolution from its CAD model and a single perspective image is proposed. Geometric properties of object of revolution are used in order to simplify the localization problem. The axis projection is first computed by a prediction verification scheme. It enables to compute a virtual image in which the contours are symmetric. A rough localization is done in this virtual image and then improved by an iterative process. Experiments with real images prove its robustness and its capability to deal with partially occluded contours.

Cite

Text

Glachet et al. "Finding the Pose of an Object of Revolution." European Conference on Computer Vision, 1992. doi:10.1007/3-540-55426-2_74

Markdown

[Glachet et al. "Finding the Pose of an Object of Revolution." European Conference on Computer Vision, 1992.](https://mlanthology.org/eccv/1992/glachet1992eccv-finding/) doi:10.1007/3-540-55426-2_74

BibTeX

@inproceedings{glachet1992eccv-finding,
  title     = {{Finding the Pose of an Object of Revolution}},
  author    = {Glachet, R. and Dhome, Michel and Lapresté, Jean-Thierry},
  booktitle = {European Conference on Computer Vision},
  year      = {1992},
  pages     = {681-686},
  doi       = {10.1007/3-540-55426-2_74},
  url       = {https://mlanthology.org/eccv/1992/glachet1992eccv-finding/}
}