Results of Audio-visual Winter Road Condition Sensor Prototype11th StandingInternational Road Weather Congress (2002)
After several years of research in winter road condition classification, an automated prototype has been tested. Classification is achieved with artificial neural networks based on data from either images of the road, acoustic signals of vehicles passing the sensor, or a combination of the two. Systems based on either images or signals give good results for some road condition classes, but the most reliable results are for the hybrid system. Hybrid results are reliable for icy, snowy, and wet road conditions but not for dry. Dry results can be improved with more
representative training data and/or further integration with other RWIS sensors. For days with icy, snowy or wet conditions, the classification system gives near 100% correct classification for all but 3 days during a 3-month winter period.
Publication DateJanuary, 2002
Citation InformationKevin McFall and T. Niittula. "Results of Audio-visual Winter Road Condition Sensor Prototype" 11th StandingInternational Road Weather Congress (2002)
Available at: http://works.bepress.com/kevin-mcfall/24/