ABSTRACT
To prevent the spread of the COVID-19 virus at construction sites, accounting for the surveillance and precautions imposed by the pandemic means frequent contact tracing, symptom monitoring, and PPE reminders. During the pandemic, many construction projects continued because of the need for a priority building that was under construction, such as for a healthcare project. To promote safety under unprecedented circumstances, this paper proposes an intelligent robotic surveillance system that can locate workers and identify whether they are properly wearing masks. The system leverages autonomous terrestrial robots equipped with wide-angle cameras to capture images of their surroundings and a two-dimensional laser scanner for simultaneous localization and mapping (SLAM). In this system, the robot is equipped with precise image processing in a well-trained convolutional neural network (VGG16) to recognize workplace entities, particularly workers and their masks, with 83.3% accuracy. Simultaneously, the mounted laser scanner enables the robot to generate the map of the surrounding environment based on the near-real-time Hector SLAM algorithm. In situ recognition would help track workers who improperly use masks and trigger interventions that would diminish the spread of the virus at the construction site. The proposed robotic system will non-intrusively and privately inform workers to use proper protocol to protect them from COVID-19 and other deadly viruses, thereby improving health and safety.