A computing platform and its tools for features extraction from on-vehicle image sequences

In the framework of the project CASSICE on the automatic classification of driving situations, the vehicle inherent dynamic parameters and the various external signals are analyzed in order to characterize the driver behavior. The objective of this project is to develop adequate tools able to generate a time-stamped database of real situations that will be used by psychologists. The driving data may be branded in several layers, starting with the lowest-level data coming form the vehicle physical sensors, which pass through many extractions phases in order to obtain structured, classifiable driving situations. The high-level data comprise information like the position of the vehicle on the road, to characterization of the road. They also include information on the other users of the road (vehicles, trucks, motorbikes, pedestrians, etc.), their characterization and their situation (relative position and speed, used lane, etc.).