Visual Human Localization and Safety Monitoring in a Digital Twin of Workspace

Digital twin plays an important role in safety monitoring of typical industrial workspaces such as substations. In the digital twin of a workspace, localizing and tracking the workers provide an intuitive way for updating their status. This paper proposes a visual human localization and safe monitoring method in a digital twin of workspace. In this paper, a 3D scanning model of the realistic workspace is utilized as its digital twin model, instead of a CAD model. The camera can be quickly calibrated with the digital twin by selecting a small number of paired 2D and 3D matching points. The pixel coordinate of human feet detected by the deep learning algorithm is mapped into the digital twin, and the human position is updated dynamically. A typical application of virtual electronic fence is developed based on the human localization results. The experimental results show that the calibration method is fast and effective, and the performance of human localization can meet the requirements of product management and safety monitoring.