Fall Prevention From Ladders Utilizing a Deep Learning-Based Height Assessment Method

According to the Center for Construction Research and Training (CPWR) Korea Occupational Safety & Health Agency (KOSHA), falls from ladders are a leading cause of fatalities. The current safety inspection process enforce height-related rules is manual time-consuming. It requires physical presence manager, whom it sometimes impossible monitor an entire area in which being used. Deep learning-based computer vision technology has potential capture large amount useful information digital image. Therefore, this paper presents deep height assessment method using single known value image measure working height, compliance rules, ensure worker safety. proposed comprises (1) extraction KOSHA database related A-type ladder; (2) object detection (Single Shot Multibox Detector SSD) (3) height-computing module (HCM) estimate (how high ground); (4) classification behavior (using developed SSD-based HCM) based on best practices derived database. algorithm been tested four different scenarios with heights ranging under 1.2 m over 2 m. Additionally, was evaluated 300 images binary (safe unsafe) achieved overall accuracy 85.33%, verifying its feasibility intelligent estimation monitoring.