Real-Time Vision-Based Warning System for Prevention of Collisions between Workers and Heavy Equipment

Abstract

The heavy equipment used at construction sites poses a significant threat to the safety of surrounding workers due to the inherently poor visibility of the equipment operator. To improve visibility, heavy-equipment manufacturers have adopted a system that monitors an all-around view of the area surrounding the equipment via cameras installed on every side (i.e., front, right, left, and rear) of the equipment body to display the surrounding environment on the operator’s monitor. Although these systems improve the visibility of the surrounding environment, detecting potential collisions with workers nearby is still restricted by the limited cognitive capacity of the operator, who is executing tasks that are themselves cognitively effortful. The aim of this study is to propose a real-time warning system using visual data acquired from cameras that are readily available in the heavy equipment to protect the workers from potentially dangerous situations involving equipment operations. For this purpose, possible collisions with workers in the workspace of heavy equipment are detected and monitored by estimating the workers’ positions in three dimensions (3D) with a monocular camera on each side of the equipment. Field tests were conducted to verify the accuracy and speed of the system, as well as its applicability to actual construction sites. The proposed system was implemented on two different sizes of excavators while the excavators performed excavating and moving tasks at various construction sites. The field test results indicate that the proposed system can provide information to the operator in real time about whether one or more workers may have come into contact with the equipment during manipulation and transportation of the equipment. It is expected that the proposed method can be utilized in around-view monitoring systems to assist the operator and achieve active safety.