Development of Visibility Assessment Methods with Digital Images Under Foggy Conditions

Abstract

Basic methods are proposed for assessing poor visibility by using digital images from a road monitoring camera. An experiment was performed over 5 days in 2002 under artificial fog at a laboratory of the National Institute for Land and Infrastructure Management in Tsukuba, Japan. Twenty-one Landolt circles were prepared as targets, each with a different variation of Landolt circle (7 circle sizes X 3 circle-contrast levels). Digital image (8 bits, red-green-blue), illuminance, and transmissivity under artificial fog were recorded together while five subjects assessed the visibility of the circle. First, the digital contrast of the Landolt circle was examined in a digital image in terms of its applicability as a visibility assessment value. Second, the average intensity of grayscale (AIG) of the entire cutout image was examined in terms of its applicability as a visibility assessment value. The magnitude of AIG represented the difference in digital intensities within the image. Digital contrast and AIG were compared with the subjective visibility assessment values (SVAV) given by the subjects. During fog, both digital contrast and AIG increased as transmissivity increased. Relationships were found between both these values and SVAV. However, when digital contrast was used, the relationships depended on circle size. When AIG was used, the relationships depended on SVAV and not on circle size or circle-contrast level. These results suggest that AIG might be more appropriate than digital contrast for assessing subjective visibility level during fog.