Fusion based safety application for pedestrian detection with danger estimation

Road safety applications require the most reliable data. In recent years data fusion is becoming one of the main technologies for Advance Driver Assistant Systems (ADAS) to overcome the limitations of isolated use of the available sensors and to fulfil demanding safety requirements. In this paper a real application of data fusion for road safety for pedestrian detection is presented. Two sets of automobile-emplaced sensors are used to detect pedestrians in urban environments, a laser scanner and a stereovision system. Both systems are mounted in the automobile research platform IVVI 2.0 to test the algorithms in real situations. The different safety issues necessary to develop this fusion application are described. Context information such as velocity and GPS information is also used to provide danger estimation for the detected pedestrians.