Publikationen

Abstract

Tunnel surveys making use of photogrammetric three-dimensional (3D) tunnel reconstruction reach resolutions in the millimeter range. Classical big data visualization approaches display point clouds only, neglecting this considerable resolution difference between structure and texture. The article suggests a data structure that separates structural and textural resolution by a regular grid on the unwrapped design surface for 3D, combined with a UV mapping technique as regularly used in computer graphics. For real-time rendering of huge multiscale data sets, the result of photogrammetric commercial-off-the-shelf reconstructions is transformed into a proprietary hierarchical data structure. It facilitates to only load currently relevant parts of the tunnel surface from the hard drive, and only upload and render currently adequate levels-of-detail onto the graphics card for seamless exploration of high-resolution geometric and image 3D tunnel data of arbitrary length. The solution allows for smooth interactive analysis and annotation such as crack identification and mapping, inventory, deformation assessment, and dimensional measurements. Aspects of data generation are addressed and information is given about the data structure, showing examples from entire tunnel 3D representations to demonstrate the smooth behaviour of the real-time rendering of huge data volumes in various scales on standard graphics hardware.