Topological Labelling of Scene Using Background/Foreground Separation and Epipolar Geometry

Abstract

The robust Principal Component Analysis (rPCA) efficiently separates an image into the foreground and background regions. The stixels provide middle-level expression of a scene using vertical columnar-superpixels of pixels with same depth computed from a pair of stereo image. Combining the classification of pixels by rPCA and depth map, topological labelling of pixels of each frame in an image sequence is achieved. The algorithm constructs static stixels and moving boxes of an image sequence from background and foreground regions, respectively. The algorithm also estimates free-space for motion planning from background regions as a collection of horizontal columnar-superpixels parallel to the epipolar lines.

Cite

Text

Hiraoka and Imiya. "Topological Labelling of Scene Using Background/Foreground Separation and Epipolar Geometry." IEEE/CVF International Conference on Computer Vision Workshops, 2019. doi:10.1109/ICCVW.2019.00079

Markdown

[Hiraoka and Imiya. "Topological Labelling of Scene Using Background/Foreground Separation and Epipolar Geometry." IEEE/CVF International Conference on Computer Vision Workshops, 2019.](https://mlanthology.org/iccvw/2019/hiraoka2019iccvw-topological/) doi:10.1109/ICCVW.2019.00079

BibTeX

@inproceedings{hiraoka2019iccvw-topological,
  title     = {{Topological Labelling of Scene Using Background/Foreground Separation and Epipolar Geometry}},
  author    = {Hiraoka, Hiroki and Imiya, Atsushi},
  booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
  year      = {2019},
  pages     = {652-660},
  doi       = {10.1109/ICCVW.2019.00079},
  url       = {https://mlanthology.org/iccvw/2019/hiraoka2019iccvw-topological/}
}