Video-based event recognition: activity representation and probabilistic recognition methods

Abstract

We present a new representation and recognition method for human activities. An activity is considered to be composed of action threads, each thread being executed by a single actor. A single-thread action is represented by a stochastic finite automaton of event states, which are recognized from the characteristics of the trajectory and shape of moving blob of the actor using Bayesian methods. A multi-agent event is composed of several action threads related by temporal constraints. Multi-agent events are recognized by propagating the constraints and likelihood of event threads in a temporal logic network. We present results on real-world data and performance characterization on perturbed data. 2004 Elsevier Inc. All rights reserved.

DOI: 10.1016/j.cviu.2004.02.005

24 Figures and Tables

0204060'04'05'06'07'08'09'10'11'12'13'14'15'16'17
Citations per Year

386 Citations

Semantic Scholar estimates that this publication has 386 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@article{Hongeng2004VideobasedER, title={Video-based event recognition: activity representation and probabilistic recognition methods}, author={Somboon Hongeng and Ramakant Nevatia and François Br{\'e}mond}, journal={Computer Vision and Image Understanding}, year={2004}, volume={96}, pages={129-162} }