Video to BrIM: Automated 3D As-Built Documentation of Bridges

Abstract

Labeling and mapping existing infrastructure is one of the grand challenges for civil engineers in the 21st century. The challenge remains in finding an appropriate data collection method that rapidly collects structural and geometrical information of bridges and automatically converts it into three-dimensional (3D) models. Over the last decade, there has been a tremendous effort to develop prototypes for capturing the as-is condition of a bridge and converting it into 3D bridge-information models (BrIMs). Bridge-information models are 3D-information-rich data models that can be used in various phases of bridge design, construction, operations, and maintenance. As an alternative solution, researchers are now studying the processing of videos to capture the required 3D information (videogrammetry) and using algorithms to automatically extract 3D objects and convert the results into a BrIM. This study evaluates the applicability of a novel videogrammetric pipeline for automatically documenting the physical condition of bridges. It also describes the results of three case studies of highway bridge assessment and modeling in which 3D information is extracted from two-dimensional video frames and point cloud data (PCD) are generated. As the next step, the PCD can be converted into a data-rich BrIM object for further processing.