Image public
Portfolio categories
The geophysical interpretation of some key subsurface geological horizons can significantly change through time, due to different interpreters and new or reprocessed geophysical data. As such, the resulting structural maps can change accordingly. Retracing and understanding such evolution in the interpretation is very valuable for any geologist new to the field. Unfortunately, it can be peculiarly time-consuming as the various structural maps are scattered in a large number of unstructured documents of diverse types. Automatic extraction of these maps, together with the associated information, can improve the efficiency of operational workflows and help professionals save time for more advanced and higher-value activities in their daily work.
Our main objective is to present an end-to-end workflow to automate the extraction of images of interest and the associated information in geoscience documents, and its application on an operational case-study provided by an energy resources company.
We developed an integrated workflow to automate the extraction of images of interest and the associated information in geoscience documents. The implemented workflow relies on a combination of free Python packages for Natural Language Processing, Computer Vision, Optical Character Recognition and Machine Learning. This workflow was applied on a case study using data from an oil field operated by Vermilion Energy, in South-West of France.
More than 90% of the relevant images and related information were extracted from the database. Besides, the whole automated process lasted only 5 hours instead of several weeks of manual work. This illustrates the enormous gains in operational efficiency which can be unlocked with such approaches, further increased by the possibility to explore the information extracted with an interactive web interface.
The proposed workflow is not limited to the use case presented, and it can be applied with minimal adjustment to any kind of information extraction within large knowledge bases.
Key features and benefits
The images of interest are extracted from the document collections and easily classified
Once all the images are extracted from the document collections, our workflow allows to further identify those that are structural maps from those that are not. This process uses a binary image classifier in a supervised Machine Learning strategy. Then, an Optical Character Recognition (OCR) is employed to extract the information associated to these images, such as the name of the interpreted structural map and its age of creation.
An interactive web application to visualize the images and explore them in a timeline
With such application, one can easily visualize on a timeline the number of structural maps associated to different reservoir horizons. In a few clicks, the user can filter on a specific time range or reservoir, browse the corresponding maps and access the source documents. The open-source Python library Plotly Dash was selected to build the app as it natively includes pre-built widgets for statistical graphs with a high level of interactivity (e.g. filtering capabilities).
Explore other demo projects related to
Join TELLUS Share and...
- Access detailed technical information on the TELLUS TOOLS prototypes, benefit from live demos and open discussions
- Listen, drive and follow IFPEN initiatives on the digital transformation of geosciences
- Receive quarterly newsletters for worldwide scientific intelligence on this fast-paced field
- ...
Check out all benefits of TELLUS Share membership
Digital transformation
towards geosciences 4.0
