LiDAR-Based Pavement Marking Detection for Construction Engineering: A Systematic Review of Methods and Applications

Abstract

Accurate detection of pavement markings is increasingly vital for construction engineering applications such as roadway asset management, site planning, progress tracking, and quality assurance. This systematic review consolidates computational methods for light detection and ranging (LiDAR)-based pavement marking detection, focusing on their relevance to automated infrastructure inspection. Using a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided methodology, 116 peer-reviewed studies are analyzed and categorized into five algorithmic paradigms: geometric, statistical, traditional machine learning, deep learning, and hybrid approaches. Hybrid pipelines, combining interpretable geometric models with data-driven learning, emerge as the most resilient to challenges common in construction environments, including occlusion, worn markings, surface clutter, and changing lighting conditions. While deep learning techniques demonstrate high accuracy, they are often constrained by large annotation requirements and computational overhead, limiting their feasibility in field-based or resource-constrained deployments. Key implementation challenges include LiDAR intensity variation, environmental noise, and scene complexity. Mitigation strategies such as radiometric correction and intensity normalization, morphology-based filtering, and temporal fusion are reviewed in depth. The paper identifies critical research gaps, such as the lack of standardized LiDAR-only benchmarks, and outlines future priorities, including domain-adaptive models, multisensor fusion, and lightweight architectures for real-time deployment in construction contexts. This review offers construction engineers actionable insights into method selection and integration for field-ready pavement marking detection pipelines.