Reconstruction of 3D structures using under-ice AUV video footage has been inaccurate, although it is not certain whether this is due to shortcomings in the methods or an inherent deficiency of the capture system. We test four different models for reconstruction on the same Antarctic under-ice AUV video (Icefin, Kamb Ice Stream): classical feature matching SfM, differentiable photometric splatting, volumetric depth fusion, and a pose-free feed-forward foundation model. All four fail, each in a distinct way: near-total registration yielding almost no structure, geometrically broken output despite healthy photometric metrics, a persistent coverage artifact, and unstable cross-chunk scale. We show that all four are consistent with the same capture regime, a small camera-baseline-to-scene-depth ratio. Following our analysis, we propose a detection system that flags this type of failure in minutes, rather than hours of wasted compute.
Figures & tables
Figure 1: TSDF fusion of box1 after increasing voxel size and truncation distance, reducing the surface from 15.0M to 667K vertices ( 22× consolidation). The fan-like silhouette persists, indicating a limitation of viewpoint overlap rather than fusion parameters.
Figure 2: VGGT on the 62-frame box6 sequence. Chunks are locally coherent (a) but merge inconsistently, with scale factors 0.43 – 1.73× (b). The recovered trajectory (c) advances predominantly toward the surface. Per-chunk baseline-to-depth ratios (d), with the 7% and 12% tier thresholds marked (dashed).
Faculty of Mathematics & Computer Science, Babes,-Bolyai University, Cluj-Napoca, Romania · Institute of Visual & Analytic Computing, University of Rostock, Germany
Department of Computer Science & Engineering and the Minnesota Robotics Institute (MnRI), University of Minnesota–Twin Cities, Minneapolis, MN 55455 USA