People Modelling and Tracking for Personal Identification

Project outline and aims

CCTV surveillance has many applications including safety, security and crowd control. The usefulness of CCTV would be increased significantly by reliable methods for identifying individuals who appear in the image sequences. The task in this project is to construct 3D models for individuals in CCTV images, including body shape, appearance (e.g. clothing) and especially the face. The models should be small for efficient storage and computation, but they must include all the information relevant for recognising the individual in other CCTV images or locating the individual in a database of models.

This project is a China Joint Project, funded by the Royal Society of London. The Chinese partner is the National Laboratory of Pattern Recognition (NLPR), which is part of the Institute of Automation, Beijing, PR China. The project began in April 2002 and will run for three years.

Project publications

· Weiming Hu, Tieniu Tan, Liang Wang and S.J. Maybank 2004 A survey on visual surveillance of object motion and behaviours. Accepted by IEEE Transactions on Systems, Man, and Cybernetics.

· Weiming Hu, Xuejuan Xiao, Dan Xie, Tieniu Tan and S.J. Maybank 2004 Traffic accident prediction using 3D model based vehicle tracking. Accepted by IEEE Transactions on Vehicular Technology.

· Weiming Hu, Dan Xie, Tieniu Tan and S.J. Maybank 2004 Learning activity patterns using fuzzy self-organizing neural network. Accepted by IEEE Transactions on Systems, Man, and Cybernetics.

· Jianguang Lou, Tieniu Tan, Hao Yang, Weiming Hu, and S.J. Maybank 2004 3D model based vehicle tracking. Submitted for publication.

· Min Hu, Weiming Hu, Tieniu Tan, Jianguang Lou and Steve Maybank 2004 Principal axis-based multi-camera correspondence for people tracking. Submitted for publication.