Lane Detection With Conditions of Rain and Night Illumination Using Hough Transform

Abstract

Lane detection is a technique to detect lane lines or lane areas that are captured from cameras or LiDAR on autonomous vehicles. At this time, the thing that still becomes a challenge in lane detection is how to detect lanes in various conditions to let autonomous vehicles detect the lanes correctly. This is because various variables can affect the final lane detection results, such as fog, rain, illumination variations, and partial occlusion. This research will discuss the lane detection method in rainy and night conditions using hough transform. There are six steps to detect lanes in rainy and night conditions using hough transform: image conversion from RGB to grayscale, noise reduction with a median filter, contrast enhancement with Contrast Limited Adaptive Histogram Equalization (CLAHE), canny edge detection, selection with Region of Interest (RoI), and line detection with hough transform. The proposed method successfully detects lane lines with an accuracy of 85.9% in 1553 video frames in rainy conditions and 946 video frames in night conditions.