Automated vision-based hardhat-wearing detection system for construction safety applications -

Abstract

The construction industry is still considered among the most dangerous industries in the world as workers are exposed to a constant risk of getting injured from falls, slips, trips, or getting struck by moving or falling objects. In an attempt to provide a safe working environment, safety programs have been trying to impose an array of provisions and regulations, of which enforcing the use of personal protective equipment (PPE), in particular hardhats, proved to be of paramount importance. As the awareness and perception of construction workers on safety and hardhat use cannot be fully trusted, the responsibility has been traditionally put into the hands of safety officers to ensure compliance with these safety regulations. However, the task of actively supervising a large construction site with a sizeable number of workers is considered tedious, inefficient, and time-consuming. Hence, this research aims at creating a vision-based system that can automatically detect a failure to wear the hardhat in videos captured from construction sites. The objective of this study is thereby two-fold: (1) evaluating existing computer vision techniques in efficiently detecting hardhats on jobsites, and (2) developing an integrated vision-based framework that can actively identify mobile construction workers then search for the presence of a hardhat in the upper region of the detected personnel. Components of the complete framework were implemented and results highlighted the potential of the proposed automated hardhat detection system in enhancing construction safety inspections.