Abstract
Aim: This paper critically reviews the application of computer vision (CV) and machine learning (ML) in construction management. Focus was on construction productivity, construction safety, construction resource monitoring, and deep learning techniques for construction progress assessment.
Methods: The study employed a critical narrative review of published literature on the application of CV and ML in construction management. The review was organized around key themes, including productivity assessment, safety monitoring, construction progress monitoring, resource and activity recognition, and disruption management.
Results: The reviewed studies demonstrate a growing application of CV and ML for worker and equipment monitoring using object detection, image classification, pose estimation, tracking, and activity recognition. These technologies have also been applied to personal protective equipment (PPE) and hazard detection, productivity assessment, and construction progress monitoring. However, the literature identifies several challenges that constrain reliable field deployment, including limited and insufficiently diverse construction datasets, occlusion, changing site conditions, poor model generalizability, difficulties in BIM integration, and privacy concerns associated with worker monitoring.
Conclusion: CV and ML have considerable potential to complement traditional construction monitoring by improving the automation and efficiency of monitoring activities. However, current systems require greater robustness and contextual adaptability to achieve reliable performance under the variable conditions encountered on construction sites.
Recommendation: Future research should prioritize the development of diverse and well-annotated construction datasets, models that are robust to occlusion and environmental changes, automated integration of CV/ML systems with construction schedules and BIM, and privacy-preserving approaches for worker monitoring.