Deep Learning Machine Vision Solutions for Monitoring Safety Structures to Supplement Building Information Models

Abstract

Building Information Models (BIM), provide a static view on building structures on design and construction time. During construction time, the safety structures and their positioning among conditions are vital for the safety of the employees on site. The models usually lack the temporospatial information regarding non-static safety structures as maintaining of such information manually is laborious. Thus, automated and continuous safety structure monitoring is costefficient and vital keeping the model up-to-date while improving construction site safety and risk information sharing among the construction related personnel during the project. This paper researches of providing up-to-date and supplementary safety structure information to building models by means of various deep learning-based machine vision solutions. The machine vision tasks consist of deep learning safety related object detection in images, point-clouds, and further instance segmentation to enable safety structure fitness determination by more traditional machine vision means