Automated construction monitoring based on computer vision: A comprehensive review

This review synthesizes recent advancements in automated construction monitoring, focusing on key dimensions including equipment, methodologies, datasets, evaluation metrics, and practical applications. It examines diverse data collection setups including single camera, unmanned aerial vehicles (UAVs), mobile phones, and multi-cameras, along with a range of models such as deep learning models and simulation-based models. The analysis highlights the critical role of dataset scale, diversity, and realism in model robustness, and reviews commonly used metrics like accuracy, precision, mean Average Precision (mAP), and Frames per Second (FPS) to evaluate performance trade-offs. Applications span safety monitoring, equipment tracking, productivity analysis, and structural health assessment. The review identifies gaps in dataset generalizability, metric standardization, and real-world validation, offering recommendations such as developing hybrid models, large-scale construction-specific datasets, and integrated multi-functional platforms. This work aims to guide future research and support the practical adoption of intelligent monitoring systems for safer and more efficient construction management.