VENUS: A System for Novelty Detection in Video Streams with Learning

May 2004

Proceedings of the Seventeenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2004)

Volume

Proceedings of the Seventeenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2004)

All Papers

Novelty detection in video is a rapidly developing application domain within computer vision. The motivation behind this paper is a learning based framework for detecting novelty within video. Since, humans have a general understanding about their environment and possess a sense of distinction between what is normal and abnormal about the environment based on our prior experience; any aspect of the scene that does not fit into this definition of normalcy tends to be labeled as a novel event. In this paper, we propose a computational learning based framework for novelty detection and provide the experimental evidence to describe the results obtained by this framework. To begin with the framework extracts low-level features from scenes, based on the focus of attention theory and then combines unsupervised learning techniques such as clustering with habituation theory to emulate the cognitive aspect of learning.