Development of Segmentation Technology for Fall Risk Areas in Small-Scale Construction Sites Based on Bird's-eye-view Images.

Construction sites have shown the highest incidence of safety accidents across industries in recent times. Small-scale sites, in particular, often operate without on-site safety managers, leading to significant safety oversights. In this study, we developed a method of identifying risk areas during construction procedures by using bird's-eye-view image data throughout the construction cycle. Actual construction site images were collected and specific target objects were selected to create an AI training dataset. The segmentation model's performance was validated, and a system was developed to identify fall risk areas by establishing interconnections between these target objects within the images. The findings of this study can help enhance compliance assessment with construction procedures and improve safety management oversight at small-scale construction sites. [ABSTRACT FROM AUTHOR]