ABSTRACT
Leveraging computer vision to automate construction progress monitoring and enhance construction safety has the potential to facilitate efficient infrastructure delivery. Despite the potentials of this area, awareness on advances remains low. Using a six-stage review method, this study examines the importance of computer vision in building infrastructure better, the challenges, current approaches and future directions necessary to deliver the technology effectively and efficiently. Findings were discussed based on 15 papers retrieved from Scopus and Web of Science database. The findings illustrate that significant research challenges encompass comprehensive scene comprehension, fluctuating tracking precision influenced by camera positioning, and the recognition of actions involving multiple equipment and workers. Challenges include absence of task-specific and measurable metrics for assessing extracted safety-related information, technical impediments stemming from the dynamic nature of construction sites, and privacy concerns. These challenges underscore the necessity for additional research in these domains.
