Advanced Unsupervised Classification Methods to Detect Anomalies on Earthen Levees Using Polarimetric SAR Imagery

Area of Science:

• Remote Sensing
• Geophysics
• Civil Engineering

Background:

• Earthen levees are critical for flood control but susceptible to slope instability caused by water dynamics.
• Early detection of levee anomalies is essential to prevent catastrophic failures.
• Traditional direct assessment methods are time-consuming and labor-intensive.

Purpose of the Study:

• To evaluate the effectiveness of fully polarimetric Synthetic Aperture Radar (polSAR) for detecting anomalies on earthen levees.
• To compare various unsupervised classification algorithms using L-band SAR data for anomaly identification.

Main Methods:

• Utilized L-band Synthetic Aperture Radar (SAR) data from NASA's UAVSAR.
• Employed unsupervised classification algorithms based on polarimetric parameters: entropy (H), anisotropy (A), alpha (α), and eigenvalues (λ).
• Applied classification techniques including H/α, H/A, A/α, Wishart H/α, Wishart H/A/α, and H/α/λ.

Main Results:

• Demonstrated the capability of L-band polSAR data to identify problematic areas on earthen levees.
• Showcased the effectiveness of unsupervised classification algorithms in detecting anomalies.
• Validated the approach using quad-polarimetric SAR imagery over the Mississippi River valley.

Conclusions:

• L-band polSAR is a viable remote sensing tool for screening levees and detecting potential instability.
• Unsupervised classification algorithms provide an efficient method for anomaly identification in levee systems.
• This technology can significantly enhance the proactive management and safety of critical flood control infrastructure.