Water extraction from SAR images based on improved geodesic active contour

Area of Science:

• Remote Sensing
• Geographic Information Systems (GIS)
• Image Processing

Background:

• Accurate water body information is vital for water resource management, flood monitoring, and ecological protection.
• Traditional methods for water boundary extraction often face challenges in accuracy and efficiency.

Purpose of the Study:

• To develop and validate an improved geodesic active contour model (IMGAC) for precise water boundary extraction from single-polarization SAR images.
• To enhance the accuracy and computational efficiency of water body detection compared to existing models.

Main Methods:

• Utilized an improved geodesic active contour model (IMGAC) incorporating an adaptive threshold for initial water body segmentation.
• Introduced a narrowband model and the signed pressure force (SPF) function into the geodesic active contour (GAC) model for iterative boundary refinement.
• Employed the F-measure and kappa coefficient to quantitatively evaluate extraction accuracy.

Main Results:

• The IMGAC model demonstrated superior performance in water boundary detection accuracy.
• Achieved a highest extraction accuracy of 96.43% using the F-measure index.
• Obtained a kappa coefficient of 0.929 and an F-measure index of 96.20% for the IMGAC method.

Conclusions:

• The proposed IMGAC model significantly improves the accuracy and efficiency of water boundary extraction from SAR images.
• This method offers a valuable reference for optimizing water extraction and boundary delineation in environmental and resource management applications.