The main obstacle to the robustness of car tracking is large shadows of vehicles. Even with a good foreground model, the tracking process is liable to be disrupted by the shadows. The paper proposes an HMM based foreground-background segmentation method which is capable of modeling shadow as well as foreground and background regions. Unlike many other probabilistic background models, it is not necessary to select the training data since the distributions for different regions can be learnt from an ordinary video sequence. The ambiguity among different regions (categories) is reduced by not only using the temporal continuity constraint for each category, but using jointly intensity and Sobel values, which measure the homogeneity of a region in space, as the observation symbols. This method itself functions a low level tracker. It can be also used as a low level process for an active contour based car tracker. Results on real-world motorway sequences show that using the proposed method, it is possible to accurately segment the image into background, cars and the shadow of cars in real time.
HMM-based segmentation of background, object and shadow from traffic monitoring movies
Titel japanisch
Transactions of the Information Processing Society of Japan ; 42 , 1 ; 1-15
2001
15 Seiten, 17 Quellen
Aufsatz (Zeitschrift)
Japanisch
Non-parametric Background and Shadow Modeling for Object Detection
Springer Verlag | 2007
|Joint segmentation and tracking of object surfaces in depth movies along human/robot manipulations
BASE | 2013
|Joint segmentation and tracking of object surfaces in depth movies along human/robot manipulations
BASE | 2013
|Adaptive background modeling with shadow suppression
IEEE | 2003
|Adaptive Background Modeling with Shadow Suppression
British Library Conference Proceedings | 2003
|