Sign in Subscribe
Machine learning

Deep Learning-based Lineament Extraction: A Comparative Study of Sentinel-1 and Landsat 9 Imagery

Strategic Insights

— 2 min read

Scientists at the Water Research and Technology Center in Tunisia have proposed an integrated approach combining convolutional neural networks (CNNs) with a support vector machine (SVM) classifier to enhance the automatic detection of geological lineaments from satellite imagery. This research study reveals that Sentinel-1, with its 10-meter radar resolution, detects a significantly higher number of lineaments (over 6 million segments) compared to Landsat 9. The results demonstrate the complementarity of the two sensors, with Sentinel-1 being more effective in detecting fine and fragmented structures, while Landsat 9 promotes spatial coherence of larger lineaments. The study's findings have the potential to improve structural mapping and pave the way for broader applications in geosciences and natural resource management.

Key Takeaways:

  • The integrated approach combining CNNs and SVMs has improved the detection of geological lineaments from satellite imagery, with Sentinel-1 detecting over 6 million lineaments compared to Landsat 9.
  • The comparative analysis highlights the complementarity of the two sensors, with Sentinel-1 being more effective in detecting fine and fragmented structures, while Landsat 9 promotes spatial coherence of larger lineaments.
  • The study demonstrates the added value of artificial intelligence and multisource integration for improving structural mapping in geosciences and natural resource management.
  • The research proposes an integrated approach combining CNNs with a support vector machine (SVM) classifier to enhance the automatic detection of geological lineaments from satellite imagery.
  • The study validates the results using Jaccard, Kappa, and coincidence indices, as well as spatial agreement with geological faults and hydrographic networks, confirming the robustness of the approach.
  • The research has been peer-reviewed and published in the journal Earth Science Informatics.
  • Additional authors for this research include Rihab Riahi and Noman Rebai.
  • The study has the potential to improve structural mapping and pave the way for broader applications in geosciences and natural resource management.

Statistics:

  • Over 6 million lineaments were detected using Sentinel-1, while Landsat 9 detected significantly fewer.
  • Sentinel-1 has a 10-meter radar resolution, while Landsat 9 has a higher spatial resolution.
  • The integrated approach combining CNNs and SVMs improved the detection of geological lineaments by 20% compared to traditional methods.
  • The study used Sentinel-1 and Landsat 9 multisource images to enhance the detection of lineaments.
  • The research was conducted by Sonia Gannouni and colleagues at the Water Research and Technology Center in Tunisia.

Sources:

  • "Deep Learning-based Lineament Extraction: a Comparative Study of Sentinel-1, Landsat 9 Imagery." Earth Science Informatics, 2025;18(3).
  • Sonia Gannouni, Water Research and Technology Center Certe, Georesources Lab, Technopk Borj Cedria, Tourist Route Soliman Nabeul, Soliman, Tunisia.
  • Rihab Riahi and Noamen Rebai, additional authors for this research.
  • Springer Heidelberg, Tiergartenstrasse 17, D-69121 Heidelberg, Germany.
  • (Springer - www.springer.com; Earth Science Informatics - www.springerlink.com/content/1865-0473/)

Read more