컴퓨터비전 기반 건설근로자 안전모 착용 여부 인식을 위한 딥러닝 기법의 적용

Abstract

Since the construction site is exposed to the external environment and works, the accident rate is higher as many risk factors are higher than other industries, and it is required to wear the Personal Protective Equipment. However, safety helmets are often not worn or worn properly in real-life situations due to frustration and annoyance. As a result, small accidents are likely to lead to serious accidents and safety managers provide safety education, but it is difficult to manage many workers in real time. Therefore, more effective safety management measures are needed. On the other hand, computer vision technology is a field of artificial intelligence, which programs computers that automatically recognize and track objects through images obtained from imaging devices. Also, the recent introduction of deep learning algorithms has been studied in many applications because the accuracy of this technology is very good. In this study, the computer vision-based deep learning object detection algorithm, Faster R-CNN, is used to supplement a person's vision and cognitive skills to learn from image data obtained from cameras installed at a construction site and to recognize whether workers wear safety hats.