CONDENSATION—Conditional Density Propagation forVisual Tracking
International Journal of Computer Vision
A Bayesian Computer Vision System for Modeling Human Interactions
IEEE Transactions on Pattern Analysis and Machine Intelligence
Activity Recognition and Abnormality Detection with the Switching Hidden Semi-Markov Model
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Hi-index | 0.00 |
The explosion in the number of cameras surveilling the environment in recent years is generating a need for systems capable of analysing video streams for important events. This paper outlines a system for detecting noteworthy behaviours (from a security or surveillance perspective) which does not involve the enumeration of the event sequences of all possible activities of interest. Instead the focus is on calculating a measure of the abnormality of the action taking place. This raises the need for a low complexity tracking algorithm robust to the noise artefacts present in video surveillance systems. The tracking technique described herein achieves this goal by using a "future history" buffer of images and so delaying the classification and tracking of objects by the time quantum which is the buffer size. This allows disambiguation of noise blobs and facilitates classification in the case of occlusions and disappearance of people due to lighting, failures in the background model etc.