Driving Perception in Challenging Road Scenarios: An Empirical Study

Vision-based road lane detection is a critical technology for autonomous driving, enabling vehicles to navigate safely and efficiently under constrained conditions such as accurately identifying the lane markings and tracking the vehicle's position. However, despite exceeding 90% detection recall in large scale datasets, existing lane detection methods often fail in some real- world scenarios such as adverse weather, intensive shadows, complex road types, and challenging lighting conditions. This study highlights these gaps and proposes potential research areas to address them. To this end, the YOLO-based lane detection algorithm is utilized as a case study for determining potential perception problems under complex traffic situations.