An Image Registration Method for Multisource High-Resolution Remote Sensing Images for Earthquake Disaster Assessment

Area of Science:

• Geospatial Information Science
• Remote Sensing Technology
• Disaster Management

Background:

• Multisource image registration is crucial for remote sensing (RS) based earthquake disaster assessment.
• Severe earthquakes cause significant deformation in RS images, challenging traditional registration methods.
• Existing methods often lack the accuracy and efficiency needed for post-earthquake analysis.

Purpose of the Study:

• To develop an improved image registration method for multisource, high-resolution remote sensing images.
• To enhance the accuracy and efficiency of registering images acquired before and after severe earthquakes.
• To provide a more reliable tool for earthquake disaster assessment.

Main Methods:

• A novel approach combining Shi-Tomasi corner detection and scale-invariant feature transform (SIFT) for tie point detection.
• An image partitioning strategy incorporating geographic information constraints.
• Outlier and redundant tie point removal using Random Sample Consensus (RANSAC) and greedy algorithms.
• Utilizing a pre-earthquake RS image database for reference.

Main Results:

• The proposed method demonstrated higher accuracy compared to classic SIFT.
• It achieved a reduced running time, improving efficiency.
• More tie points were detected with a more even distribution across the images.
• Performance was validated on image pairs from regions affected by severe earthquakes.

Conclusions:

• The developed method significantly improves the accuracy and efficiency of remote sensing image registration for post-earthquake scenarios.
• It effectively handles image deformation caused by severe seismic events.
• This advancement offers a more robust solution for earthquake disaster assessment using remote sensing data.