Visual inspection

Process (computing)

Engineering

Tilt (camera)

Statistics

Computer security

3D reconstruction

Operating system

Bridge (graph theory)

Computer science

Geology

Structural health monitoring

Outlier

SAFER

Point cloud

Mathematics

Image (mathematics)

Internal medicine

Artificial intelligence

Structural engineering

Matching (statistics)

Medicine

Forensic engineering

Noise (video)

Computer vision

Journal of Bridge Engineering

Imagery-based, three-dimensional (3D) reconstruction from unmanned aerial vehicles (UAVs) holds the potential to provide safer, more economical, and less disruptive bridge inspection. In support of those efforts, this paper proposes a process using an imagery-based point cloud. First, a bridge inspection procedure is introduced, including data acquisition, 3D reconstruction, data quality evaluation, and subsequent damage detection. Next, evaluation mechanisms are proposed including checking data coverage, analyzing point distribution, assessing outlier noise, and measuring geometric accuracy. The overall approach is illustrated in the form of a case study with a low-cost UAV. Areas of particular coverage difficulty involved slim features such as railings, in which obtaining sufficient features for image matching proved challenging. Shadowing and large tilt angles hid or weakened texturing surfaces, which also interfered with the matching process.

UAV Bridge Inspection through Evaluated 3D Reconstructions