Iterative Robust Graph for Unsupervised Change Detection of Heterogeneous Remote Sensing Images

Area of Science:

• Remote Sensing
• Computer Vision
• Image Analysis

Background:

• Heterogeneous remote sensing images pose challenges for change detection due to differing imaging mechanisms.
• Direct comparison of heterogeneous data is often unreliable, necessitating advanced methods.

Purpose of the Study:

• To develop a robust graph mapping approach for unsupervised heterogeneous change detection.
• To overcome the limitations of direct comparison for images from different modalities.

Main Methods:

• Constructing K-nearest neighbor graphs to represent image structure.
• Utilizing graph mapping to compare structures and generate difference images, invariant to imaging modality.
• Employing a Markovian co-segmentation model with co-graph cut for change detection.
• Iteratively refining graph construction using detected changes to enhance robustness.

Main Results:

• The proposed method effectively detects changes in heterogeneous remote sensing imagery.
• Experimental results on diverse datasets validate the approach's effectiveness.
• The iterative framework improves graph robustness and final detection performance.

Conclusions:

• The graph mapping approach offers a robust solution for unsupervised heterogeneous change detection.
• The method successfully leverages invariant structural information for reliable change identification.
• The developed technique enhances the accuracy and reliability of remote sensing change detection.