Carbonate Rock Fracture Identification Method Based on an Improved YOLOv5 Algorithm

• Published in: Pure and applied geophysics Vol. 181; no. 1; pp. 189 - 201
• Main Authors: Xie, Jun, Gao, Renjie, Zhang, Yuanpei, Zhang, Jianguo, Xia, Yong, He, Yilin
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 2024; Springer Nature B.V
• Subjects: Accuracy; Acoustics; Algorithms; Carbonate rocks; Carbonates; Data logging; Earth and Environmental Science; Earth Sciences; Fractured reservoirs; Fractures; Geophysics; Geophysics/Geodesy; Identification; Identification methods; Logging; Methods; Object recognition; Oil and gas fields; Petroleum; Fracture recognition; deep learning; YOLOv5 algorithm
• ISSN: 0033-4553; 1420-9136
• DOI: 10.1007/s00024-023-03408-6

Fractures play a crucial role in discovering and developing petroleum reserves. However, traditional logging techniques face significant challenges in identifying fractures. To address such challenges, this article proposes a new method that combines conventional logging data with a small amount of marker data from cores and image logs to identify fracture development in reservoirs with high accuracy and fast operation. The proposed method is based on the improved YOLOv5, which offers a new idea for fracture identification. The fracture data from the carbonate rocks of the Sulige gas field in the Ordos Basin were used for training and validation. Finally, positive experimental outcomes were achieved, demonstrating the usefulness of the improved YOLOv5 algorithm in detecting fracture development in carbonate rocks.