Estimating Iceberg Melt Rates from Video Using Novel View Synthesis

Understanding iceberg melt is crucial for assessing climate change effects in polar regions. This paper investigates the geometric reconstruction and volume estimation of icebergs using multi-view imagery, combining Structure-from-Motion (SfM) and Neural Radiance Fields (NeRFs). High-resolution multi-view datasets are processed through SfM techniques to estimate camera poses, which are subsequently used to generate a 3D model of the object using NeRF. The proposed framework successfully reconstructs iceberg geometry and estimates its volume, ultimately providing estimates of melting rates. This study demonstrates the possibility of applying SfM and NeRFs for monitoring iceberg dynamics, offering a new approach to analyzing iceberg volume and volume flux.