Developing a near infrared based night vision system

Accidents occurring at night represent a significant part of the total number of road fatalities. Thus, intelligent systems for supporting the driver during night promise to have a significant impact on traffic safety. In particular, the second generation of such systems provide semantic, symbolic information in order to effectively draw the driver's attention towards the actual danger source. This paper describes the design choices and the system architecture of the EDEL project. Moreover, we present the algorithms we use for detection of pedestrians, which are frequently involved in accidents at night, and discuss early results. In particular, the implemented system provides recognition rates similar to the state of the art in daylight conditions, even if this requires a more in-depth analysis of the captured scene, with a consequent performance penalty in terms of response time.