Retrieval from Video and Pictorial Databases Employing Similarity and Motion
Abstract
This project focuses on research towards the development of a system for similarity based retrieval from video and pictorial databases. The specific aims of the work are development of an image understanding system for detecting humans in videos, development of an expressive query language called Hierarchical Temporal Logic (HTL), for spatio-temporal queries on video databases and an initial development of activity recognition module that tracks and recognizes human actions in videos. In this project, we have already developed head and face detection algorithms and also developed an initial tracking system for human body parts. An initial graphical base query system was developed for expressing spatio-temporal queries on video databases. All these algorithms are now combined in a web demo. Our plans for the coming six months which are left in this 1 year program is to extend our tracking, recognition and querying algorithms. We plan to develop more robust head and limb detection and also to develop an initial algorithm for human activity recognition as well as extending the querying system to include hierarchical query structures.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1999
- Accession Number
- ADA398364
Entities
People
- A. P. Sistla
- Clement Yu
- Jezekiel Ben-arie
Organizations
- University of Illinois at Chicago