A Study on the Detection of Protective Helmets for the Safety of Construction Workers

Abstract

In recent years, with the rapid development of the construction industry, the number of accidents and deaths on construction sites increased, so the prevention of accidents is one of the important issues. Worker safety during construction is a major concern of the construction industry. Wearing helmets can reduce injuries among construction workers, but helmets are not always worn and used correctly for a variety of reasons. Therefore, computer vision-based automatic helmet detection systems are very important. Although many researchers have developed machine and deep learning-based motorcycle helmet detection systems, there is little research on helmet detection for construction workers. Therefore, in this research work, an automatic system for detecting the helmets of construction workers based on real-time computer vision is presented. In this study, machine learning method is used to detect helmets, and a model is trained using 1,500 images. The test results show that the average accuracy is above 95% in laboratory conditions.