Surveillance event detection

Mert Dikmen, Huazhong Ning, Dennis J. Lin, Liangliang Cao, Vuong Le, Shen Fu Tsai, Kai Hsiang Lin, Zhen Li, Jianchao Yang, Thomas S. Huang, Fengjun Lv, Wei Xu, Ming Yang, Kai Yu, Zhao Zhao, Guangyu Zhu, Yihong Gong

Research output: Contribution to conferencePaperpeer-review


We have developed and evaluated three generalized systems for event detection. The first system is a simple brute force search method, where each space-time location in the video is evaluated by a binary decision rule on whether it contains the event or not. The second system is build on top of a head tracker to avoid costly brute force searching. The decision stage is a combination of state of the art feature extractors and classifiers. Our third system has a probabilistic framework. From the observations, the pose of the people are estimated and used to determine the presence of event. Finally we introduce two ad-hoc methods that were designed to specifically detect OpposingFlow and TakePicture events. The results are promising as we are able to get good results on several event categories, while for all events we have gained valuable insights and experience.

Original languageEnglish (US)
StatePublished - 2008
Externally publishedYes
EventTREC Video Retrieval Evaluation, TRECVID 2008 - Gaithersburg, MD, United States
Duration: Nov 17 2008Nov 18 2008


ConferenceTREC Video Retrieval Evaluation, TRECVID 2008
Country/TerritoryUnited States
CityGaithersburg, MD

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Vision and Pattern Recognition
  • Human-Computer Interaction
  • Software


Dive into the research topics of 'Surveillance event detection'. Together they form a unique fingerprint.

Cite this