Rigorous analysis of safety rules for vision intelligence-based monitoring at construction jobsites
Doyeop Lee, Numan Khan, Chansik Park
Abstract
Construction safety rules play a vital role in mitigating accidents and fatalities at construction sites. Many researchers are currently devoted to monitoring the rule compliance using computer vision-based approaches; however, such systems are still not mature for application at construction job sites. A single autonomous source’s job-site safety control such as CCTV is a nontrivial task that requires a detailed analysis of safety rules for a work-stage based compact vision intelligence system. This paper proposes a grounded theory methodology (GTM) to systematically classify the safety rules for practical implementation using vision intelligence technology. The rules are classified into four groups employing open coding, axial coding, and selective coding approaches: (1) before work, (2) with intervals, (3) after work, and (4) during work. The proposed GTM-based model is further linked with scene-capturing sources, such as single-scene-capturing through smartphones for the required rules: (a) before work and (b) after work, (c) periodic scene-capturing using robots and drones for the required rules with intervals, and (d) CCTVs for the safety rules to continuously monitor safety. It is anticipated that this research will pave the new direction in computer vision intelligence-based construction safety management.