Sensor-based Proactive Workers’ Safety Status Evaluation on Construction Site

Abstract

The construction industry has an unsatisfactory safety performance. Employing approximately 7% of the world’s workforce, the construction industry accounts for 30-40% of all workplace fatalities. The causes of these fatalities fall predominantly among the “Fatal Four”-falls, electrocution, struck by objects and caught-in-betweens. Previous research shows that about 80% of the accidents are strongly associated with workers’ unsafe behaviors. An evaluation of construction workers’ safety behavior seems in order and is the focus of the current study. The biggest challenge of implementing a proactive and automatic behaviors related safety risk evaluation is to fill the gap between the computer systems’ objective interpretation and human mind’s subjective understanding of the workers’ behaviors. Then how can the workers’ behaviors on the construction sites be quantitively represented is the key.

In this study, a proactive sensor-based workers’ safety evaluation framework is proposed. Two characteristics of construction workers’ behaviors are identified as the position and posture. By measuring these two features, researchers can quantitively analyze the safety risks of the behaviors. The proactive evaluation framework, based on the current sensor technologies, consists of two main parts: (1) real-time workers’ position monitoring and (2) automatic posture identification during construction work. In daily life, localization technology is considered to have matured; when applying localization systems to construction sites, however, two persistent challenges are deployment efficiency and stability due to their dynamics and high complexity of construction environment. Consequently, a multi-sources fusion-based indoor localization algorithm is specifically designed for the construction job site environment. The performance of the proposed algorithm is evaluated in two testbeds. Furthermore, a quantitative model is developed to evaluate the impact of a specific localization system on the safety decision-making performance. This model can be utilized by the safety management team to facilitate their selection of an appropriate localization system for their application. For the automatic posture identification, instead of human body silhouette detection, a 3D human body skeleton is employed as the input of the posture identification. Doing so makes the detection background and situation independent, thus, making the detection is more stable and flexible. State-of-the-art human body skeleton-capturing technologies are employed, making the posture detection more practicable. To encode the detected skeletons during a worker’s action, a body part relative position-based posture-coding scheme is proposed. The coding enables automatic interpretation of the postures with semantic information. Then the safety experts’ knowledge of safety situation evaluation can be transferred into the proactive behavior-based safety management framework.

The core contribution of this work is that the position and posture are identified can be utilized as the quantitive indicators to represent workers’ behaviors, then proposed a framework to fuse the position and posture information to evaluate the worker’s safety status on construction sites. This research makes it possible to monitor, automatically and proactively, the safety behaviors of construction workers. The results of this work offer four useful outcomes: 1) a prototype framework for implementing proactive workers’ safety status evaluation; 2) a practical and flexible indoor localization application in the construction industry; 3) a quantitative tools for evaluating the impact of indoor localization systems on safety monitoring performance; and 4) a positive human body posture identification model with semantic information. Moreover, this study offers a feasible approach to transfer safety experts’ knowledge into computer systems, and to eventually contribute to safety management in the construction industry.

Date of Award	8 Sept 2018
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Xiaowei LUO (Supervisor)