Vision-Based Detection of Unsafe Actions of a Construction Worker: Case Study of Ladder Climbing

Abstract

About 80-90% of accidents are caused by the unsafe actions and behaviors of employees in construction. Behavior management thus plays a key role in enhancing safety, and particularly, behavior observation is the most critical element for modifying workers' behavior in a safe manner. However, there is a lack of practical methods to measure workers' behavior in construction. To analyze workers' actions, this paper uses an advanced and economical depth sensor to collect motion data and then investigates consequent motion-analysis techniques to detect the unsafe actions of workers, which is the main focus of this paper. First, motion data are transformed onto a three-dimensional (3D) space as a preprocess, motion classification is performed to identify a typical prior, and the selected prior is used to detect the same action in a testing data set. As a case study, motion data for unsafe actions in ladder climbing (i.e.,backward-facing climbing, climbing with an object, and reaching far to a side) are collected and used to detect the actions in a new testing data set in which the actions are randomly taken. The result shows that 90.91% of unsafe actions are correctly detected in the experiment.