In this paper a system for the analysis of color images of traffic scenes is presented. This system, which is based on a suitable processing of the chromatic information contents of the images, aims at recovering in a robust way important parts of a road scene like the road-bed, obstacles which are present on the road-bed, horizontal and vertical road signs, and traffic lights. A wide experimentation on images of urban and country roads has shown the power of using color information compared to the performance we obtained by algorithms for object detection and recognition that rely solely on geometric features extracted from monochromatic images. The results suggest that color cue can be very helpful to construct drivers' assistant systems based on computer vision technologies even in the presence of quite complex road environments.
Color cues for traffic scene analysis / De Micheli, E; Prevete, Roberto; Piccioli, G; Campani, M.. - STAMPA. - (1995), pp. 466-471. (Intervento presentato al convegno Intelligent Vehicles '95 Symposium tenutosi a Detroit, USA nel September 25-26, 1995) [10.1109/IVS.1995.528338].
Color cues for traffic scene analysis
PREVETE, ROBERTO;
1995
Abstract
In this paper a system for the analysis of color images of traffic scenes is presented. This system, which is based on a suitable processing of the chromatic information contents of the images, aims at recovering in a robust way important parts of a road scene like the road-bed, obstacles which are present on the road-bed, horizontal and vertical road signs, and traffic lights. A wide experimentation on images of urban and country roads has shown the power of using color information compared to the performance we obtained by algorithms for object detection and recognition that rely solely on geometric features extracted from monochromatic images. The results suggest that color cue can be very helpful to construct drivers' assistant systems based on computer vision technologies even in the presence of quite complex road environments.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.