Robust Lane Departure Warning System for ADAS on Highways

Abstract

Lane detection in Advanced Driver Assistance System (ADAS) has a significant role and must be handled well. Lane detection is the core for Lane Departure Warning System (LDWS) that warns the driver upon unintendedly leaving the lane. Unfortunately, road type conditions such as straight and curved road, weather conditions such as fog, low illumination can increase the difficulty of finding the lane lines. In this paper, lane departure warning system that is based on image processing technique for detecting lane lines is implemented. Video frames are fed to the pre-processing phase where brightness and contrast of images are adjusted, then Red Green Blue (RGB) images are changed to Hue Saturation Lightness (HSL) colour space. Gaussian blur filter to remove noise. Sobel edge detection to detect and filter lane lines. Perspective warp to convert view to flat 2D surface. Sliding window technique to differentiate between the left and right lane boundaries. Finally, after fitting polynomial equations that represents lane boundaries, the vehicle offset from lane center can be calculated. The algorithm is tested using self-made dataset as well as on Raspberry pi 4 on highway in Egypt that includes daytime, nighttime, and foggy video frames, Results showed that it needs improvements. © 2022 IEEE.