Automated monitoring and systemic analysis of workers' safety behavior in construction operations

Han, Sang Uk

Permalink

https://hdl.handle.net/2142/45601 Copy

Description

Title

Automated monitoring and systemic analysis of workers' safety behavior in construction operations

Author(s)

Han, Sang Uk

Issue Date

2013-08-22T16:49:15Z

Director of Research (if dissertation) or Advisor (if thesis)

Pena-Mora, Feniosky A.

Doctoral Committee Chair(s)

Liu, Liang Y.

Committee Member(s)

• Pena-Mora, Feniosky A.
• Lee, SangHyun
• El-Rayes, Khaled A.
• Golparvar-Fard, Mani

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Date of Ingest

2013-08-22T16:49:15Z

Keyword(s)

• Construction management
• safety
• behavioral safety
• motion capture and recognition
• simulation
• complex systems

Abstract

In construction, workforces are an essential element of safety management. Previous studies found that about 80%–90% of accidents are strongly associated with workers’ unsafe actions and errors, which are affected by safety-related factors (e.g., safety program, safety culture, production pressure). Behavior measurement also turns out to be significantly correlated to safety outcomes (e.g., incident rates). To this end, an in-depth understanding of workers’ behavior has tremendous potential to contribute to the prevention of construction injuries and illnesses. However, there are immense difficulties regarding worker observation in construction, which is the single most significant element in managing workers’ behavior. These difficulties are three-fold: (1) field observation is time-consuming, painstaking, and error-prone due to the complexity of workforce and workplace on a jobsite; (2) training workers for observation is expensive and inefficient for the short tenure of construction workers, particularly hired by various subcontractors; and (3) workers’ active participation is required for peer-to-peer observation, survey, and interview. In an effort to provide a robust and automated means for worker observation, this study proposes a framework of vision-based unsafe posture and action detection for human behavior monitoring. The framework consists of: (1) the identification of critical unsafe actions and postures; (2) the collection of relevant motion templates (i.e., identified critical unsafe actions); (3) the extraction of a 3D human skeleton model from videos collected with a stereo camera system (e.g., a 3D camcorder, multiple cameras); and (4) the detection of unsafe actions using the motion templates and 3D skeleton models. As a case study, experimental studies were undertaken to evaluate each process of the framework. This research thus opens up the possibility of micro-level motion tracking and recognition with video cameras to identify the frequency and types of workers’ unsafe actions on jobsites. The resulting information will serve as: (1) preliminary information for providing workers with direct feedback on their behavior; (2) a positive safety performance measurement to evaluate ongoing safety management; and (3) data to analyze the impact of safety-related factors (e.g., schedule pressure, rework, safety training) on workers’ behavior. In particular, the proposed approach will allow one to provide feedback on workers’ safety behavior, to foster communications between a manager and workers through the feedback process, and eventually to contribute to the creation of a safe climate on a jobsite.

Graduation Semester

2013-08

Permalink

http://hdl.handle.net/2142/45601

Copyright and License Information

Copyright 2013 Sang Uk Han

Owning Collections