Ranking Geological Cross-Sections for database querying

Amandine Fratani and Sophie Viseur and Fabrice Popineau and Pierre Henry and Badih Ghattas and Georges Oppenheim and Damien Dhont and Claude Gout. ( 2021 )
in: 2021 RING Meeting, ASGA

Abstract

Geological cross-sections are convenient supports for synthesizing geological knowledge of given geological subsurface structures. They contain information about the geological time of the formations and their architecture. Querying a database of case study reports or papers with geological cross-sections could be then a convenient way to find geological analogues. Image-content query is used to achieve this kind of query. It is often performed using deep-learning, but these techniques require numerous pairs of original images and associated human-tagged images in order to obtain satisfying results after the training phase. Alternative approaches rely on image similarity measures based on textures, colours or shapes contained in the images. In this paper, we propose such a method for ranking geological cross-sections using two kinds of similarity measures: (i) the presence and proportion of formations from similar geological times; (ii) the global geological architecture. This approach combines the use of a colour dictionary and different correlation measures. Noddy is a software allowing synthetic geological cross-sections to be generated using a series of geological events (faulting, tilting, folding, etc.). A database of 100,000 geological cross-sections was generated using Noddy. In each cross-section, the layers are colored following the standard ICS color codes. Results of cross-section rankings are shown from this database and discussed.

Download / Links

BibTeX Reference

@inproceedings{FRATANI_RM2021,
 abstract = { Geological cross-sections are convenient supports for synthesizing geological knowledge of given geological subsurface structures. They contain information about the geological time of the formations and their architecture. Querying a database of case study reports or papers with geological cross-sections could be then a convenient way to find geological analogues. Image-content query is used to achieve this kind of query. It is often performed using deep-learning, but these techniques require numerous pairs of original images and associated human-tagged images in order to obtain satisfying results after the training phase. Alternative approaches rely on image similarity measures based on textures, colours or shapes contained in the images. In this paper, we propose such a method for ranking geological cross-sections using two kinds of similarity measures: (i) the presence and proportion of formations from similar geological times; (ii) the global geological architecture. This approach combines the use of a colour dictionary and different correlation measures. Noddy is a software allowing synthetic geological cross-sections to be generated using a series of geological events (faulting, tilting, folding, etc.). A database of 100,000 geological cross-sections was generated using Noddy. In each cross-section, the layers are colored following the standard ICS color codes. Results of cross-section rankings are shown from this database and discussed. },
 author = { Fratani, Amandine AND Viseur, Sophie AND Popineau, Fabrice AND Henry, Pierre AND Ghattas, Badih AND Oppenheim, Georges AND Dhont, Damien AND Gout, Claude },
 booktitle = { 2021 RING Meeting },
 publisher = { ASGA },
 title = { Ranking Geological Cross-Sections for database querying },
 year = { 2021 }
}