AUTHOR=Wang Meng , Zhou Quan , Yin Lu , Dong Yu 

TITLE=Automatic detection and classification of igneous rock fractures from imaging logging using K-mean algorithm and DNN

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 13 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2025.1640215

DOI=10.3389/feart.2025.1640215

ISSN=2296-6463

ABSTRACT=Accurately identifying fracture zones and their types in strata is of great significance for enhancing oil and gas recovery efficiency. Due to its complicated geological structure and long-term weathering and erosion, the buried hill reservoir in Huizhou Oilfield has developed a complicated reservoir structure. This structure is characterized by great burial depth, strong heterogeneity, diverse lithological types, and high degrees of weathering. These factors collectively result in significant spatial variability in fracture development patterns, making fracture identification and classification a highly challenging task. To address this challenge, this study proposes a fracture identification method based on image segmentation and recognition technology using electrical imaging logging. The method first employs the K-means clustering algorithm combined with morphological processing to segment electrical imaging logging images, thereby optimizing sample quality and improving the accuracy of fracture information extraction. Subsequently, a deep neural network is introduced for fracture structure recognition, fully leveraging the advantages of deep learning in pattern recognition and feature extraction to achieve highly accurate fracture detection. Especially under small-sample conditions, this approach effectively enhances recognition performance. Finally, fracture characteristic parameters are extracted to classify the reservoirs, allowing for the selection of high-quality reservoirs and laying the foundation for improved recovery rates. In practical application of the model, this method successfully identified dissolution fractures, semi-open fractures, and continuous fractures within the samples, verifying its effectiveness in detecting different types of fractures. Through high-precision image processing techniques, the identification accuracy was effectively ensured, providing more precise geological interpretation and technical support for the drilling and development of the buried hill reservoir in Huizhou Oilfield.