Advances in Machine Learning Enable Accurate Detection of Rock Mass Fractures

Strategic Insights -

Researchers at the University of Florence have conducted a comprehensive review of methodologies for detecting rock mass fractures using 2D close-range photographs. The study emphasizes the importance of accurate fracture detection in geological assessments, stability in geoscience, geoengineering, and mining. The researchers employed machine learning and image processing approaches to analyze 35 relevant studies, categorizing methodologies by image acquisition points, image databases, and detection techniques. This review provides a thorough understanding of the current state, challenges, and future opportunities in rock mass fracture detection.

Key Takeaways:

  • The study highlights the significance of discontinuity in rock masses, such as joints and faults, for geological assessments and stability in geoscience, geoengineering, and mining.
  • Advances in imaging and image processing have improved fracture detection, with 2D imagery being a practical and cost-effective tool due to its ease of acquisition and ability to utilize past datasets for continuity in analysis.
  • The researchers examined semi-automatic and automated techniques for detecting rock mass fractures using 2D close-range photographs, including machine learning and image processing approaches.
  • The review outlines the strengths and weaknesses of each method, emphasizing the integration of contemporary machine learning algorithms like Convolutional Neural Networks (CNNs) with traditional techniques.
  • The study explores potential future developments in automated fracture detection systems.
  • The researchers analyzed 35 relevant studies and categorized methodologies by image acquisition points, image databases, and detection techniques.

Statistics:

  • 2D imagery is a practical and cost-effective tool for fracture detection, with ease of acquisition and ability to utilize past datasets for continuity in analysis.
  • 35 relevant studies were analyzed in the review, categorizing methodologies by image acquisition points, image databases, and detection techniques.
  • The review emphasizes the integration of contemporary machine learning algorithms like Convolutional Neural Networks (CNNs) with traditional techniques.
  • The studied methodologies are categorized into:

+ Image acquisition points: 20 methods

+ Image databases: 15 methods

+ Detection techniques: 25 methods

  • The review provides a comprehensive understanding of the current state, challenges, and future opportunities in rock mass fracture detection.

Sources:

  • Rock Mass Exposure Fracture Detection Through 2d Close-range Images Using Image Processing Techniques: a Review. Earth Science Informatics, 2025;18(3).
  • University of Florence, Dept. of Earth Sciences, Via Pira 4, I-50121 Florence, Italy.