| Title | Machine learning applied to geoscience: Geo-referenced character recognition |
| Download | Download (whole publication) |
| |
| Licence | Please note the adoption of the Open Government Licence - Canada
supersedes any previous licences. |
| Author | Griffiths, M ;
Russell, H A J; Logan, C E |
| Source | Southern Ontario groundwater project 2014-2019: summary report; by Russell, H A J (ed.); Kjarsgaard, B A (ed.); Geological Survey of Canada, Open File 8536, 2020 p. 65-74, https://doi.org/10.4095/321092  Open Access |
| Year | 2020 |
| Publisher | Natural Resources Canada |
| Document | open file |
| Lang. | English |
| Media | on-line; digital |
| Related | This publication is contained in Southern Ontario
groundwater project 2014-2019: summary report |
| File format | pdf |
| Subjects | general geology; Science and Technology; modelling; models; lithostratigraphy; bedrock geology; lithology; sedimentary rocks; evaporites; salt; sandstones; shales; limestones; carbonates; dolostones;
bedrock topography; software; boreholes; oil wells; gas wells; bathymetry; geophysical logging; core samples; unconformities; groundwater; cuestas; Leapfrog Works; Port Lambton Group; Kettle Point Formation; Hamilton Group; Marcellus Formation;
Dundee Formation; Lucas Formation; Onondaga Formation; Amherstburg Formation; Sylvania Formation; Bois Blanc Formation; Bass Islands Formation; Salina Group; Guelph Formation; Lockport Group; Clinton Formation; Cataract Formation; Queenston
Formation; Georgian Bay Formation; Blue Mountain Formation; Trenton Group; Black River Group; Shadow Lake Formation; Findlay Arch; Algonquin Arch; Frontenac Arch; Methodology; Artificial intelligence; machine learning; Geographic data; Automation;
Phanerozoic; Paleozoic; Devonian; Silurian; Ordovician; Cambrian |
| Illustrations | sketch maps; flow diagrams; tables; schematic representations; graphs |
| Program | Groundwater Geoscience Aquifer Assessment & support to mapping |
| Released | 2020 05 28 |
| Abstract | Significant quantities of non-digital geoscience data exists on maps. In many cases this information has been scanned and is available in a raster format, but remains irretrievable for digital
operations. This information may be in both a text and symbol format and it is also necessary to capture the georeferenced location. In many cases this data may also consist of handwritten characters, which have much greater variability than typed
characters. An example of such a dataset is handwritten depth soundings that are a common aspect of Canadian Hydrographic Service (CHS) field sheets. CHS maintains a collection of scanned and georeferenced digital image files with handwritten depth
soundings recorded directly on lake maps. To make use of this data for digital 3-D modelling, it needed to be converted to geo-referenced vector data. To avoid the time-consuming process of entering thousands of data points, a machine-learning
algorithm was applied to automate the digitization process using open-source software. Robust machine-learning libraries available in Python were integrated within a custom work environment for this application. This is an example of how analogue
geoscience datasets can be captured in a cost effective, timely and reliable manner. |
| Summary | (Plain Language Summary, not published)
Collection of papers on work completed in the past five years as part of the southern Ontario Groundwater Project. This edited volume is a collection of
currently unreported work. |
| GEOSCAN ID | 321092 |
|
|