ML Anthology

Morse Functions for Activity Classification Using Spatiotemporal Volumes

Naresh P. Cuntoor
CVPRW 2006 pp. 20
doi:10.1109/CVPRW.2006.129 /cvprw/2006/cuntoor2006cvprw-morse/

Abstract

Activities can be represented using spatiotemporal volumes or xyt cubes that are formed by stacking frames in a video sequence. Objects moving in the scene carve out patterns in the volume. In this paper we present an algebraic way of characterizing the topology of the volume using Morse theory. We propose a Morse function based on dynamics of moving objects and use the critical points and the associated critical values as a signature of the activity. Experiments using the TSA airport surveillance dataset are used to demonstrate the usefulness of the proposed method for activity recognition.

CVPRW Semantic Scholar

Cite

Text

Cuntoor. "Morse Functions for Activity Classification Using Spatiotemporal Volumes." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2006. doi:10.1109/CVPRW.2006.129

Markdown

[Cuntoor. "Morse Functions for Activity Classification Using Spatiotemporal Volumes." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2006.](https://mlanthology.org/cvprw/2006/cuntoor2006cvprw-morse/) doi:10.1109/CVPRW.2006.129

BibTeX

@inproceedings{cuntoor2006cvprw-morse,
  title     = {{Morse Functions for Activity Classification Using Spatiotemporal Volumes}},
  author    = {Cuntoor, Naresh P.},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2006},
  pages     = {20},
  doi       = {10.1109/CVPRW.2006.129},
  url       = {https://mlanthology.org/cvprw/2006/cuntoor2006cvprw-morse/}
}