A Computer Vision Approach for Lane Detection and Tracking

Abstract

According to the World Health Organization, each year about 1.35 million people are involved in road accidents. These mishaps bring great economic losses to victims, their families, and nations. A system for guiding the drivers is one among the foremost important aspects of recent vehicles; to ensure the security of the driver and to scale back the danger of auto accidents on the road. The proposed system helps the driver to spot the lane through which he or she is meant to drive, and if the vehicle deviates from the right lane, then an alert is shown on the screen. The screen also displays the radius of curvature at each point consistent with the road, along with the position of the vehicle depending on the middle of the lane. The proposed system uses a camera attached to the dashboard of the vehicle which makes it easy for the camera to record the video of the road ahead. This system isn't only helpful for the drivers to avoid accidents but can also be employed by the automated vehicles for following certain lanes. The multiple advantages of this technique emphasize the very fact that this is often a promising system. The expenses for this technique are very low considering the fact that the majority of the vehicles now accompany a camera mounted on its front. This system considers two algorithms; the first one makes use of Canny edge detection along with Hough transform, and the next one is based on Sobel operator along with perspective transform. The latter came out to be more accurate and precise in detecting the lanes along with the detection of curves in these lanes. Hence, for developing the final system the Sobel operator and the perspective transform were used.