Bilkent University
Department of Computer Engineering


Object Classification and Human Action Recognition in Video


Yigithan Dedeoglu
Ph.D Student
Computer Engineering
Bilkent University

Video surveillance has long been in use to monitor security sensitive areas such as banks, department stores, highways, crowded public places and borders. The increase in the number of cameras in ordinary surveillance systems overloaded both the human operators and the storage devices with high volumes of data and made it infeasible to ensure proper monitoring of sensitive areas for long times. In order to filter out redundant information generated by an array of cameras, and increase the response time to forensic events, assisting human operators with the identification of important events in video by the use of "smart" video surveillance systems has become a critical requirement. In this presentation an instance based machine learning algorithm and system for real-time object classification and human action recognition which can help to build intelligent surveillance systems will be addressed. The proposed method makes use of object silhouettes to classify objects and actions of humans present in a scene monitored by a stationary camera. An adaptive background subtraction model is used for object segmentation. Template matching based supervised learning method is adopted to classify objects into classes like human, vehicle, and human group; and human actions into predefined classes like walking, running, hand waiving, boxing, kicking and falling by making use of object silhouettes.


DATE: April 24, 2006, Monday@ 15:40