Safety analysis and localization of scaffold joints using uav-acquired images

A scaffold is one of the structures that causes many casualties at construction sites, so it must be monitored frequently. However, since scaffold monitoring is currently performed through visual inspection, it is time-consuming and costly. In particular, scaffold joints are difficult to monitor frequently due to their small size and large number. For efficient joint monitoring, this study proposes an automated monitoring method using an unmanned aerial vehicle (UAV). The proposed method consists of three stages: (1) detecting joints in UAV-acquired images using a deep learning-based object detection algorithm, and identify and match detections of the same joint using point clouds generated by simultaneous localization and mapping (SLAM); (2) analyzing the installation status of the joint using a rule-based classifier based on the geometric characteristics of the detected components; and (3) estimating the joint locations from the SLAM point cloud and visualizing the results in a joint status map. As a result, the proposed method achieved a 100% recall and a 74.8% F1 score in classifying unsafe joints. Additionally, the locations of the joints were estimated with an accuracy of 98.4%. The experimental results confirm that the proposed approach can be effectively utilized for monitoring scaffold joints. The contributions of this study lie in proposing the first UAV-based method for monitoring scaffold joints, designing a classifier for automatically analyzing their installation status, and constructing a data set from various construction sites for detecting joints and their components.