SAR image segmentation using Voronoi tessellation and Bayesian inference applied to dark spot feature extraction

Area of Science:

• Remote Sensing
• Image Processing
• Oceanography

Background:

• Oil spills pose significant environmental risks.
• Accurate detection of oil spills from satellite imagery is crucial for mitigation efforts.
• Synthetic Aperture Radar (SAR) is a key technology for monitoring marine environments.

Purpose of the Study:

• To develop a new segmentation-based algorithm for oil spill feature extraction from SAR intensity images.
• To accurately segment dark spots and their marine surroundings.
• To extract shape and distribution parameters for improved oil spill recognition.

Main Methods:

• The algorithm integrates Voronoi tessellation, Bayesian inference, and Markov Chain Monte Carlo (MCMC) simulation.
• It segments SAR scenes into homogenous regions: dark spots and marine surroundings.
• The method was validated using both real and simulated SAR intensity images.

Main Results:

• The algorithm successfully extracts shape and distribution parameters of dark spot areas.
• Segmentation into distinct regions allows for detailed feature analysis.
• Quantitative evaluation using simulated data confirmed the algorithm's accuracy.

Conclusions:

• The proposed algorithm provides valuable features for oil spill identification.
• This method enhances the capability for automated oil spill detection using SAR data.
• The extracted parameters are beneficial for subsequent oil spill classification stages.