MONITORING MOUNTAIN CRYOSPHERE DYNAMICS BY TIME LAPSE STEREO PHOTOGRAMMETRY

Abstract. In this paper, we propose a method to monitor surface dynamics in mountain cryosphere environments, based on a device with two fixed cameras at two different locations and with a convergent angle. Computer vision methods are used to derive pixel displacements between images taken at different times and to reconstruct the 3D geometry of the scene to convert pixel displacements into meter displacements. The proposed methods overcome the drawbacks of traditional methods such as lack of time repeatability or lack of spatial resolution. We present the results on two study sites instrumented with the device located in the French Alps: a temperate glacier and a rock glacier. We propose different ways to assess the accuracy of each processing steps solely based on stereo images. The method is validated using traditional measurements (GPS and LiDAR) and shows results comparable or even more robust than these traditional methods.