Fog detection from camera images
How analysis of traffic cam images along highways can improve fog detection and prediction.
The occurrence of fog is a complicated phenomenon, it may be very local and appear very quickly. Therefore, fog detection is a challenge. The present network of automatic weather stations and visibility sensors is far from dense enough, and optical & thermal satellite images have unfavourable acquisition times and/or resolution. However, the department of public works (RWS) operates a network of 300 thousand traffic cams which have more potential to detect fog on a local scale.
In this project we cooperate with RWS to explore the possibilities to detect fog conditions in traffic cams using machine learning image classification techniques. First results using 160 traffic cams are promising. Ultimately this research should lead to detection system with direct interaction with traffic participants, to improve safety and traffic flow.