A novel Two-level Hierarchical Dirichlet Processes (HDP) model is proposed to understand dynamic traffic scenes. The model is described by the Chinese Restaurant Franchise (CRF), and used to analyze traffic surveillance video sequences which contain hierarchical patterns with complicated motions and co-occurrences. Without any prior knowledge of the traffic rules, activities are detected as distributions over moving pixel patches, while traffic phases are discovered as distributions over activities according to the traffic signals. Both activity and traffic phase numbers are automatically optimized. The results show that our model can successfully discover both activities and traffic phases which make veracious description and perception of traffic scenes.
Motion perception for traffic surveillance
2014-10-01
630605 byte
Conference paper
Electronic Resource
English
Vehicle Shape Approximation from Motion for Visual Traffic Surveillance
British Library Conference Proceedings | 2001
|NTRS | 1982