Abstract:

A novel method is proposed to study representation and recognition of traffic event based on context in this paper. Firstly, foreground moving objects and their motion parameters are extracted by video detection and tracking; and then context is divided into three classes: spatial context, temporal context and object context by the surroundings traffic event occurred in. From these work, the concept of event cell is developed by the integration of moving objects properties and contextual information. Basic event and complex event are represented according to the execution scheduling of event cells. The event recognition method is presented at last. Based on video surveillance of real scenes, the method is validated that it can be well applied to recognize traffic event automated, and it could become an efficient technique support for the management and control of intelligent transportation.

Key words: intelligence transportation, video surveillance, traffic event, context