Effects of the Wet Retroreflectivity and Luminance of Pavement Markings on Lane Departure Warning in Nighttime Continuous Rain with and without Glare Sources 2019-01-1014

A common challenge for both machine vision (MV) systems based on visible-spectrum cameras and for human drivers is detection of pavement markings in nighttime rainy conditions. This occurs because a layer of water refracts the light differently than air, causing conventional markings to substantially retroreflect the light away from the driver or camera when the marking is immersed. This reduces the marking retroreflectivity in wet conditions, and thus the contrast in the image collected by the camera at longer viewing distances. MV lane departure warning (LDW) systems also depend on pixel data from shorter viewing distances; the contrast here also depends on diffuse luminance of the marking, which typically is reduced if the marking is soiled or worn. The efficacy of LDW systems is also known to be influenced by the presence of glare, as regions of glare test the limits of the high dynamic range of the sensor and reduces the contrast between the marking and the surrounding roadway surface. In this study, the authors explored the effects of wet retroreflectivity and luminance of white and yellow markings on the detection performance of a Mobileye LDW system in continuous nighttime rain conditions supplied by a rain tunnel. Pavement marking samples were evaluated in the absence and presence of glare sources, provided by vehicle headlights. LDW performance is reported as confidence level of marking detection for the various testing conditions. The percentage of detection at a certain confidence level (e.g. high or medium and high combined) was found to correlate with a two-factor interaction of the diffuse luminance property and the continuous wet retroreflectivity. In the presence of glare, the confidence of detection of the markings correlated with this two-factor interaction, in all cases where differentiated rates of detection were observed.