Dynamic fall risk assessment framework for construction workers based on dynamic Bayesian network and computer vision

Due to the dynamics of changing construction-related entities at construction sites and the hazardous work environment, safety accidents occur frequently, especially falls from heights. The current practice of fall risk assessment for construction workers, which mainly relies on manual observation by safety experts, is a static risk assessment that is time-consuming and laborious. A proactive, dynamic risk assessment framework is urgently needed to address this issue. In this work, computer vision has been combined with dynamic Bayesian network (DBN) to propose a dynamic risk assessment framework. The aim of the proposed framework is to improve the efficiency of risk assessment and reduce fall risk by automatically detecting onsite risk factor information. The proposed framework was tested using the activity of climbing ladders as a case study. The results show that the proposed dynamic fall risk assessment framework is feasible. It can be used to dynamically assess the fall risk of workers by automatically detecting the states of fall risk factors and capturing dynamic changes among the risk factors. The framework also includes a method of sending targeted early warnings to workers while assessing their risk levels, reducing the possibility of falls.