| Summary: | This research aims to produce glacier velocity data from terrestrial time-lapse imagery using automated feature tracking. Glacier velocity measurements are key to understanding sub-glacial hydrology and the propagation of calving events. However existing programs for processing time-lapse imagery use sparse grid based tracking, which does not always fulfil the potential of the spatial data. PyTrx, a new approach, applies object-oriented design to pre-process, feature-track and interpolate time-lapse data, producing velocity rasters from multiple cameras. The program is evaluated using imagery from Kronebreen glacier, Svalbard. This reveals velocity estimates for Kronebreen of under 5m/day, consistent with other independent measurements. PyTrx also finds texture in the velocity field of Kronebreen, consistent with variable velocity due to crevassing. Identifiable inaccuracy in the velocity measurements is likely caused by poor quality of the DEM used to project to geographic space and weather affecting the ability to track features. The research suggests improvements to feature tracking in poor weather, the refinement of algorithms to eliminate noise and alternative interpolation methods as potential further developments.
|
|---|