A Visual-based Method for Safety Status Monitoring of Site Protection Facility

Abstract

The site accidents are more dangerous for workers in the construction industry than any other industry. The accidents usually occur when the protection facility fails to function due to management failure. As a result, safety monitoring plays an important role in preventing construction accidents. This study presents a visual-based method with a specialized label to be recognized effectively by computer visualization technique. The falling protection by the sudden opening of an elevator on the construction site was selected for the case study. A Faster R-CNN deep learning model was adopted for the proposed visual-based safety monitoring method. According to the testing results on a real construction site, the proposed method achieved the rates of 00% of clearness rate and 98.0% of correctness. Moreover, the proposed method required only a specialized water-proof identification tag (ID-tag) and the regular Closed-Circuit Television (CCTV), which is precise, effective, and cost-effective than other advanced methods available nowadays. The proposed method allows practical applications in the safety management of construction sites.