<![CDATA[Lane detection plays a crucial role in autonomous driving systems, as it directly affects path planning, steering control, and overall driving safety. Early solutions based on conventional computer vision techniques offered initial progress, but their sensitivity to changing lighting, weather, and road conditions restricted their reliability in real-world scenarios. With the rise of deep learning, semantic segmentation approaches have significantly enhanced detection accuracy. However, models relying on a single network structure often struggle to balance broad contextual understanding with precise spatial detail. To overcome these limitations, recent research has increasingly focused on hybrid architectures that combine multiple complementary design strategies.This review systematically examines hybrid lane detection methods that integrate elements such as segmentation and detection pipelines, attention mechanisms, multi-scale feature extraction, and temporal modeling. We introduce a structured classification of hybrid frameworks, review key representative approaches, and analyze performance patterns across commonly used benchmark datasets. Ongoing challenges—including accurate detection of thin lane markings, robustness under occlusion, and generalization across diverse driving environments—are also discussed. To provide practical insight, a case study of a context-aware and spatially focused hybrid model is presented. Finally, future research directions are explored, including transformer-based designs, multi-modal sensor fusion, and computationally efficient solutions for real-time deployment. This review seeks to unify scattered research efforts and support the development of more reliable and robust lane detection systems for autonomous vehicles.]]>
