Robust Nighttime Road Lane Line Detection using Bilateral Filter and SAGC under Challenging Conditions

Abstract

In the last two decades, Advanced Driver Assistance Systems (ADAS) has been one of the most actively conducted areas of studies for reducing traffic accidents. Road lane line detection is one of the essential modules of ADAS. Lots of advancement has been already done, but most of the recent papers did not consider the wide variability of challenging nighttime conditions. In this paper, a method to detect nighttime lane line under different challenging conditions proposed. This simple technique can reach the real-time computation for ADAS applications and at the same time, can handle multiple challenges at a time. In the beginning, Bilateral Filter has been used to reduce the noise while preserving the edges. Next, we choose an optimized threshold (OT) for the Canny edge detector, which can detect edges under a wide variability of nighttime illumination conditions. After that Region of Interest (ROI) is selected using an equilateral triangle-shaped mask which helps to reduce computation time and remove unwanted edges. After that, lines are extracted by Probabilistic Hough Transform (PHT). Finally, a robust technique Slope and Angle based Geometric Constraints (SAGC) is proposed to remove the non-lane lines extracted by PHT. SAGC reduce false detection significantly. Experimental results show that the average detection rate is 94.05%, and the average detection time is 26.11ms per frame which outperformed state-of-the-art method.