Abstract
A motion detection and tracking algorithm is presented for monitoring the pedestrians in an outdoor scene from a fixed camera. A mixture of Gaussians is used to model each pixel of the background image and thus adaptive to the dynamic scene. Colour chromaticity is used as the image representation, which results in the illumination-invariant change detection in a daylit environment. To correctly interpret those objects that are occluded, merged, split or exit from the scene, a scene model is created and the motion of each object is predicted. A Bayesian network is constructed to reason about the uncertainty in the tracking. The results for detecting and tracking the moving objects in the PETS sequences are demonstrated.
| Original language | English |
|---|---|
| Title of host publication | Proceedings - 2nd IEEE International Workshop on Performance Evaluation of Tracking and Surveillance (PETS’2001) |
| Subtitle of host publication | In conjunction with IEEE Computer Society Conference on Computer Vision and Pattern Recognition |
| Editors | James Ferryman |
| Number of pages | 8 |
| Publication status | Published - 9 Dec 2001 |
| Externally published | Yes |