DIMACS Workshop on Video Mining

November 4-6, 2002
DIMACS Center, Rutgers University, Piscataway, New Jersey

Azriel Rosenfeld, University of Maryland, ar@cfar.umd.edu
Daniel DeMenthon, University of Maryland, daniel@cfar.umd.edu
Dave Doermann, University of Maryland, doermann@cfar.umd.edu
Presented under the auspices of the Special Focus on Data Analysis and Mining.

Modern computer technology, together with the proliferation of broadcast channels and of video-based surveillance systems, has enabled us to produce vast amounts of both raw and processed video data. The potential uses of this data are many and varied. Monitoring and mining of the content of this already huge, rapidly growing mass of data calls for the development of major computational resources and the development of sophisticated video understanding techniques.

The applications and potential applications of video mining include:
- monitoring of (possibly remote) surveillance cameras in theft protection, fire protection, care of bedridden patients and young children
- automatic recognition of suspicious people in large crowds
- automatic checking to identify passengers entering an airplane, bus, or public building (to verify entry authorization)
- automatic quality control in manufacturing processes
- robotic vision
- retrieval of archived video clips to illustrate a newsbreaking story
- retrieval of suspicious activities in prerecorded video surveillance sequences
- browsing of DVD and set-top box recordings
- intelligent fast-forward techniques
- detection of multi-lingual scene text to determine origin of broadcast
- video search engines for web browsers
- ranking of video clips by relevance in web query results
- classification of videos into genres for search pruning
- detection of crowd patterns for mob control
- detection of traffic patterns for traffic understanding management
There are some pervasive challenges here. A huge amount of data must be efficiently stored. Once stored, there must exist efficient retrieval algorithms. It must also be possible to retrieve different kinds of data stored in possibly different formats. Some of the applications will require real-time response and action based on an incoming data stream.

The purpose of this workshop will be to survey available and potential technologies for video monitoring and mining (and in general methods of fast and efficient content-based analysis of video streams) and to identify promising directions for research in this challenging area. Specific topics to be covered will be analysis of camera motion and scene activity; temporal segmentation; content-based classification, indexing, and retrieval; representation, browsing, and visualization.

We will also investigate related mining problems having to do with audio mining, seismic data mining, and cross-modality mining. Audio mining applications of interest include speech recognition; seismic data mining applications include identification of potential new sources of oil and gas and detection of earthquakes and/or nuclear tests. Cross-modality issues arise for example in problems involving identification from both video and speech.

Next: Call for Participation
Workshop Index
DIMACS Homepage
Contacting the Center
Document last modified on January 4, 2002.