Advances in Artificial Intelligence for Hydrographic Feature Mapping

Strategic Insights -

Researchers from Southern Illinois University have made significant improvements in mapping hydrographic features such as river boundaries, streamlines, and waterbodies using high-resolution digital elevation models (HRDEMs) from LiDAR and InSAR technologies. According to the study, drainage crossings, which facilitate the passage of drainage flows beneath roads, are often not represented in HRDEMs, resulting in erratic or distorted hydrographic features. The study aimed to address this issue by developing advanced deep learning models to classify drainage crossings using HRDEM-derived geomorphological features.

Key Takeaways:

  • The study found that high-resolution digital elevation models (HRDEMs) from LiDAR and InSAR technologies have significantly improved the accuracies of mapping hydrographic features.
  • Drainage crossings, which facilitate the passage of drainage flows beneath roads, are often not represented in HRDEMs, resulting in erratic or distorted hydrographic features.
  • The study developed advanced CNN models, EfficientNetV2, using four co-registered 1-meter resolution geomorphological data layers derived from HRDEMs for drainage crossing classification.
  • The advanced CNN models with HRDEM, TPI (21 x 21), and a combination of HRDEM, POS, and TPI (21 x 21) improved classification accuracy in comparison to the baseline model by 3.39, 4.27, and 4.93%, respectively.
  • The study culminated in explainable artificial intelligence (XAI) for evaluating those most critical image segments responsible for characterizing drainage crossings.
  • The study included researchers Michael Edidem, Bill Xu, Ruopu Li, Di Wu, Banafsheh Rekabdar, and Guangxing Wang from Southern Illinois University.

Statistics:

  • The advanced CNN models improved classification accuracy in comparison to the baseline model by 3.39, 4.27, and 4.93%.
  • The study used 1-meter resolution geomorphological data layers derived from HRDEMs for drainage crossing classification.
  • The study included 4 co-registered geomorphological data layers derived from HRDEMs.
  • The study used a combination of HRDEM, POS, and TPI (21 x 21) to improve classification accuracy.

Sources:

  • Deep learning classification of drainage crossings based on high-resolution DEM-derived geomorphological information. Frontiers in Artificial Intelligence, 2025,8.
  • Frontiers Media S.A. (publisher)
  • VerticalNews (news reporter)