| Title | Quality of image-based manganese nodule abundance assessment |
| Download | Download (whole publication) |
| |
| Licence | Please note the adoption of the Open Government Licence - Canada
supersedes any previous licences. |
| Author | Schoening, T; Greinert, J |
| Source | Program and abstracts: 2017 GeoHab Conference, Dartmouth, Nova Scotia, Canada; by Todd, B J ; Brown, C J; Lacharité, M; Gazzola, V; McCormack, E; Geological Survey of Canada, Open File 8295, 2017 p. 108, https://doi.org/10.4095/305928  Open Access |
| Links | GeoHab 2017
|
| Year | 2017 |
| Publisher | Natural Resources Canada |
| Meeting | 2017 GeoHab: Marine Geological and Biological Habitat Mapping; Dartmouth, NS; CA; May 1-4, 2017 |
| Document | open file |
| Lang. | English |
| Media | on-line; digital |
| Related | This publication is contained in Program and abstracts: 2017
GeoHab Conference, Dartmouth, Nova Scotia, Canada |
| File format | pdf |
| Subjects | economic geology; marine geology; surficial geology/geomorphology; environmental geology; geophysics; mapping techniques; oceanography; marine environments; conservation; marine organisms; marine
ecology; resource management; biological communities; environmental studies; ecosystems; mineral deposits; manganese nodules; marine sediments; faunas; sampling methods; geophysical surveys; acoustic surveys, marine; bathymetry; bedforms; Biology;
Renewable resources |
| Illustrations | digital images |
| Program | Offshore Geoscience |
| Released | 2017 09 26 |
| Abstract | Manganese nodules are a marine mineral resource and are considered for deep sea mining operations. These nodules constitute an important element of the deep sea habitats they occur in and their
abundance and size frequencies have an impact on occurring fauna. Assessing the distribution of nodules is traditionally done with a combination of large-aerial hydro-acoustic mapping linked with ground-truthing by physical sampling. While
hydro-acoustics provide large aerial coverage (km2/h) with low resolution (m/px), physical sampling provides low aerial coverage (cm2/h) with high resolution (mm/px). To bridge these two separate data domains, optical imaging has successfully been
applied as it provides medium aerial coverage (ha/h) and resolution (cm/px).
Extracting quantitative data from optical images is traditionally done by effortful manual image annotation. More recently, multiple automated and semi-automated image
analysis algorithms have been proposed. These algorithms are usually tuned for one specific data set or use case. The application of these algorithms to other optical imagery data sets is one necessity to prove their robustness. As manual annotations
of manganese nodules are scarce and focus on nodule counts rather than exact nodule delineations, quantitative assessment of the quality of detection algorithms in the form of e.g. precision and recall is not possible at the moment.
Apart from
the within-data comparison, a link to the traditional sampling strategies is required. These strategies are the de-facto standard for aerial mapping of habitats and assessing seafloor substrate composition (including manganese nodules). In the case
of physical sampling, statistic variations in the natural nodule abundance can bias the sampling outcome. In the case of hydro-acoustic sampling, small-scale natural variations in abundance that are relevant to mining as well as habitat composition
can be occluded due to the limited resolution. Using optical imaging as a bridge technology enables to extract more robust nodule abundance data.
This presentation will include results on comparing different nodule detection algorithms, and will
show the challenges in correlating physical sampling derived data with optical imagery data and shows potential applications for habitat assessment using the presented algorithms. |
| Summary | (Plain Language Summary, not published)
The sixteenth annual GeoHab Conference was held this year (2017) at the Waterfront Campus of the Nova Scotia Community College in Dartmouth, Nova Scotia,
Canada. |
| GEOSCAN ID | 305928 |
|
|