Lane Detection for Autonomous Driving: A Comprehensive Review

Lane Detection plays a fundamental and critical role in autonomous driving systems, which can provide accurate road structure information for vehicles and lay a visual foundation for downstream trajectory prediction and planning control. Despite its significance, few papers survey existing lane detection algorithms, leading to unclear research gaps and technical challenges. To this end, this paper reviews lane detection comprehensively, ranging from datasets, loss functions and evaluation metrics to 2D and more advanced 3D lane detection, with the aim of presenting a clear and complete technical chain for developing lane detection algorithms. Specifically, the paper proposes a taxonomy for lane detection and analyzes the technical principles, advantages, and limitations of each category. Benchmark experiments are introduced to reveal the trade-off relationships between complexity and performance. Finally, we identify seven promising research directions that address current limitations in the field, charting a path toward safer, more efficient, and more reliable autonomous driving systems.