Mitigating Human-related Hazards in Construction Projects: An Agile Framework for Jobsite Safety Assessment through Integrating Multiple Data Sources

Description

Construction companies incur financial losses and negative impacts to reputation due tolabor fatalities and injuries. In Hong Kong, more than 25% of all workplace injuries andfatalities occur in the construction industry. Since more than 70% of all accidents arerelated to human activities, detecting and mitigating risks associated with humanbehavior can improve the negative public perception of the construction industry.Because construction environments are dynamic and human behavior is unpredictable,conventional observational approaches to detecting these risks have proven to be time-consumingand unreliable. Therefore, sensing technologies such as RGB-D cameras,IMUs and RFIDs have been introduced to automatically detect human motion in relationto its surrounding environment. Thus, sensing technologies can potentially address manyof the difficulties in hazard detection and safety assessment on construction sites.Integrating existing sensing tools will improve hazard detection and safety assessmentby generating data from a variety of perspectives capable of better reflecting theinterdependence of human behavior and the environment on construction sites. However,existing sensing technologies currently lack a universal data integration and processingframework, which is problematic because it results in noisy, incomplete, andincompatible data. These problems create fundamental obstacles for improving safetyassessment.This proposed research aims to develop an agile framework to: 1) accurately collectmultiple human behavior-related sensing data, 2) efficiently detect anomalies andhazards, and 3) correctly assess jobsite safety conditions. The goal of the research is todesign a generic and self-adaptive framework that can integrate new and cumulativedata sources. Specifically, the framework will: 1) define an agile protocol to collect, clean,unify and merge data from multiple sensing sources, 2) propose procedures and modelsto automatically recognize dangerous activities through temporal data segmentsprocessed according to the protocol, and 3) assess safety conditions on specific projectsand validate the proposed framework.The framework will be able to provide project managers and workers with an automatedsystem to identify, record and assess onsite hazards, which will improve the effectivenessof safety management practices in the construction industry. In turn, improvements insafety management will result in a reduction of costs related to labor injuries andfatalities, and an increase in the safety and health of construction workers in HongKong.