Dimensional Uncertainty Quantification for Laser-Scanning-Based Underwater 3D Reconstructions

High-resolution seafloor mapping from autonomous underwater vehicles (AUVs) or remotely operated vehicles is an important tool for assessing benthic ecosystems and infrastructure, which is increasingly becoming routine. While single surveys show a snapshot in time of the state of such environments, repeat surveys can identify how these change over time. To gauge whether changes are statistically significant, it is necessary to quantify the uncertainty of such 3D reconstructions; however, most 3D reconstruction methods do not provide any such measure. While there are various sources of uncertainty when reconstructing seafloor terrain data from structured light surveys, this research focusses on the contribution from the line laser pose uncertainty to the dimensional uncertainty of laser-scanning-based underwater 3D reconstructions. A Monte Carlobased approach is used to model the laser fan pose uncertainty from stereo images of the laser line projection acquired from various distances from the seafloor. The method is demonstrated on data acquired at the Southern Hydrate Ridge off the coast of Oregon with the SeaXerocks 3 mapping device deployed with the AE2000f AUV.