KIAM Main page Web Library  •  Publication Searh   

Conference material: "Proceedings of the International Conference on Computer Graphics and Vision Graphicon (19-21 September 2023, Moscow)"
Authors: Bolotov I., Tolstaya E., Mezghani M., Khalifa A.
Deep Learning for Segmentation of Cement Spots in Core Images
Detecting anhydrite spots on core images is an important task in oil&gas industry due to the impact of anhydrite on overall reservoir quality and production potential of hydrocarbon reservoirs. Accurate identification of anhydrite spots helps to differentiate zones of high and low permeability and identify areas with reduced reservoir quality. Traditional segmentation algorithms are often not accurate for anhydrite segmentation. Therefore, deep learning methods are promising to obtain stable and reliable segmentation results. However, the required training data are not always available. Within the presented research, we suggested utilizing synthetic data and preprocessing techniques to extend the training data for the task at hand. Our methodology for anhydrite segmentation relying on state-of-the-art neural networks showcased the ability to detect and delineate anhydrite spots of varying shapes and sizes on example of representative experimental data. Integrating proposed method into a comprehensive reservoir analysis could enhance the pipeline of geological analysis.
Deep learning, segmentation, anhydrite, DCRF, Unet++, FPN, DeepLabV3, StyleGAN
Publication language: english,  pages: 11 (p. 463-473)
English source text:
Export link to publication in format:   RIS    BibTeX
About authors:
  • Bolotov Ivan,,  Aramco Research Center
  • Tolstaya Ekaterina,,  Aramco Research Center
  • Mezghani Mokhles,,  EXPEC Advanced research Center Saudi Aramco
  • Khalifa Aqeel,  ,  EXPEC Advanced research Center Saudi Aramco