Absolute coregistration and doming correction in adverse conditions, or how to retrieve snow depth from drone flights

Abstract

Doming, a superimposition of a scene-wide dome-shaped bias to the elevation data, is a classical problem in photogrammetry, especially when using the auto-calibration and bundle adjustment typical of the Structure-from-Motion approach. A number of solutions have been proposed, but most rely on a certain degree of freedom in the data acquisition strategy, such as the ability to acquire many ground control points or to acquire additional imagery over the area of interest with different flight parameters. In this paper, we present a method to solve for the dome when working on a snowfield in the Norwegian mountains. That specific situation leads to difficulties to fly the drone optimally and acquire ground control points due to short weather windows, high winds, cold temperatures, low sun angles and short day duration (i.e. high latitudes). Here, we use natural landmarks and pre-existing knowledge about the bare ground topography for removing the inevitable dome..