Lane Marking Detection Using LiDAR Sensor

To achieve fully autonomous driving, the vehicle needs visualization of the surrounding environment, and this makes it dependent on multiple perception sensors. Lane detection is significant in this case, as multiple tasks rely on its accuracy, for example, Simultaneous Localization And Mapping (SLAM), automatic lane keeping and lane centring which is commonly used in Advanced Driving Assistance Systems (ADAS), and other functions that require lane departure or trajectory planning decisions. These functions are responsible for minimizing the number and severity of road accidents, as they enable the car to position itself within the road lanes properly. Lane marking is challenging to model due to the road scene variations, and therefore, it is a complicated task. In this paper, we implement an automated algorithm for extracting road markings using a LiDAR point cloud utilizing the variances intensity properties. Our technique detects lane line coordinates based on computer vision algorithms, and without any dependency or knowledge of the test field parameters, like road width or centre-line coordinates etc. Experimental testing is conducted on a test field with ground-truth coordinates of the lane markings, and it shows that the proposed algorithm provides a promising solution to the lane marking detection.

VIVO