Lane Detection Using Deep Learning for Rainy Conditions

Prior research has shown that various road marker classification mechanisms in clear or dry weather conditions have high accuracy performance. However, the performance tends to be lower under rainy driving conditions due to the reduced quality of the road image when detecting the five classes of road markers which are Single, Single-Single, Dashed, Solid-Dashed, and Dashed-Solid. To address this challenging condition, lane marker detection based on deep learning approach is proposed in this paper. The target weather condition is rainy, which is very challenging as it causes the surface of the roads, especially the area which includes the lane marker to become blurry and unclear due to the rainwater. In order to carefully select the right features of the road such that the lane marker can be classified and detected successfully. The lane marker object is captured from the frames of the video clips taken from established published video datasets. With this fast and better lane marker detection, the achievable classification precision is satisfactory although the weather is rainy.