ABSTRACT
As construction projects resume worldwide and workers return to the job site, the possibility of transmitting the Covid-19 could be added to the extensive list of risks confronting workers in the construction sites;thus, the workers need to work alone in an assigned activity. Many workers are already working alone in the construction sites, such as utility workers, repair technicians, teleworkers, operators, and drivers. Lone workers in construction are subjected to greater safety risks compared with those working alongside others. Considering the accidents faced by lone workers, it’s less likely that another person would be there to aid them - and if they don’t get treatment quickly enough, serious injuries might prove deadly. Currently, the construction sites depend on physical inspections to the construction sites and manual observation of video streams generated through close circuit television (CCTV). To solve this issue, this research work presents an automated deep learning-based fall detection system of a lone worker to provide information of severe situations and help the workers in their golden time. A diverse dataset of multiple scenarios having workers with the excavator, forklift, ladder, and mobile scaffold is established, and a deep learning algorithm has been trained to validate the concept. The developed system is expected to reduce the efforts being made in manual inspection, enhance the timely access of the due aid from co-workers and supervisors, which is more easily obtainable in non-lone working situations. © 2021 Proceedings of the International Symposium on Automation and Robotics in Construction. All rights reserved.