Ocean Floor Characterization and Change Detection through High-Resolution Photo-Mosaicing

Seafloor imagery is routinely acquired during near-bottom mapping surveys conducted with both remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs). Deep-sea hydrothermal fields or shallow-water coral reef communities are, for instance, two scenarios that have long been the target of such studies. Imagery is useful to characterize the nature and distribution of geological features and biological communities, extract ecological indicators, and to provide a permanent visual record of the seafloor condition. However, imaging studies often yield large numbers of images (several tens of thousands, especially in deep-sea cruises) that are frequently underutilized largely because of the difficulties inherent in processing and visualizing large data sets. Moreover, light suffers from a rapid and nonlinear attenuation underwater that affects the acquired images. In this talk we will describe ongoing work at the University of Girona towards development and application of vision-based seafloor survey methodologies, including large area 2D mosaicing (>1sqkm), monocular-based 3D mosaicing, and stereo seafloor modeling. The developed tools set a first step towards detecting and documenting the temporal variations associated with the active processes operating at these sites. We will also present the natural extensions of this work to automatically detect and monitor changes in the seafloor under strong 3D relief and lack of navigation data.