Bayesian Object Detection in Dynamic Scenes
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Real Time Robust Human Detection and Tracking System
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops - Volume 03
Real-time foreground-background segmentation using codebook model
Real-Time Imaging
| Hi-index | 0.00 |
The problem addressed in this work is the achievement of accurate and real-time detection of target in uncontrolled environments, which are typically characterized by dynamic background and lightning changes. Two main contributions are presented. Starting from a state-of-the-art background subtraction approach based on non-parametric codebook model [1], we first propose some algorithmic improvements leading to a reduced processing time with a slight increase of detection accuracy. Secondly, we developed a method to remove false detections caused by uncovered background areas. At the same time, this proposed algorithm contributes to eliminate most erroneously detected clusters and allows decreasing the background initialization period to a single frame. The analysis demonstrated that the proposed approach constitutes one more step toward efficient and real-time automated video monitoring system for uncontrolled environments.