Rapid Three-Dimensional Shaft Scanning from Spherical Video with Photogrammetry
Ryan Preston, Golder Associates Ltd.
Detailed three-dimensional (3D) surveying of shafts using modern scanning methods can be difficult, time consuming and potentially prone to errors accumulated over the length. Similarly, manned shaft inspections can be prone to human induced errors and have limited repeatability. Remote video inspections offer a potential solution to both these issues. A spherical camera and lighting system can be mounted directly to the skip or lowered remotely on a winch. Video data are then captured over the length of the shaft for a detailed, permanent record of conditions which can be interrogated in detail later without tying up the hoist. These video data can be processed using structure from motion based photogrammetric methods to generate 3D models of the inspected shaft length. The quality and accuracy of the models is highly dependent on data collection conditions as they impact quality of the photographs and the availability of survey control. Photogrammetry models have been shown to accurately capture the shape of objects when compared to overlapping LiDAR surveys with mean fits of 2 mm but, depend on control points within the model area for accurate scaling. Three shafts were modelled over their entire length using photogrammetric models built from video inspection data using a FLIR Ladybug® camera. Difficulty with accumulated errors were encountered in most models making them suitable for general planning and change detection, but not detailed surveying. Model generation is relatively low effort for personnel but is computationally intensive and benefits from a dedicated computer and an approximately 1-week minimum lag time between data collection and delivery.
Scanning, Photogrammetry, Video Survey