Instant Road Marking Line Segmentation Method for Highly Automated Vehicles

The paper deals with methods and means of detecting road marking lines in images obtained from cameras of highly automated vehicles (HAVS). The proposed method is based on a two-stage detection process, in which first the lines are segmented (Instance Segmentation), and then parameterized models of the lines are determined using the neural network model YOLOv8n-seg, which can be further used in algorithms of autonomous driving. Such a process allows efficient identification and localization of road marking lines in complex road conditions. In this paper, the first stage of road marking line detection, i.e., intrinsic segmentation, is considered. Test results of the proposed approach are presented, including an evaluation of its performance and accuracy on different test datasets.