Image based fog detection and visibility estimation for driving assistance systems

Abstract

An advanced driving assistance system (ADAS) must also take into account the weather conditions. One of the most dangerous weather conditions for driving scenarios is the presence of fog. So an important task for a driving assistance system is to detect the presence of fog, estimate the fog's density and determine the visibility distance of the driver. Our method is based on a single in-vehicle camera and is actually an improvement over existing fog detection solutions, in terms of speed and accuracy. We are able to detect day time fog in a wide range of scenarios, including urban scenarios. After detecting the presence of fog in the image and based on the fog's density we are able to compute the visibility distance and inform the driver about the environment's weather conditions.