A Novel Bayes Approach to Impervious Surface Extraction from High-Resolution Remote Sensing Images

Area of Science:

• Remote Sensing
• Urban Planning
• Geographic Information Systems (GIS)

Background:

• Impervious surfaces are key indicators of urbanization, essential for urban planning and management.
• Current methods often rely on medium-low-resolution imagery, limiting accuracy for dynamic urban development monitoring.
• High-resolution impervious surface extraction is needed to address shadow misclassification issues.

Purpose of the Study:

• To develop and validate a novel impervious surface extraction method for high-resolution satellite images.
• To address the challenge of shadow misclassification in urban remote sensing data.
• To evaluate the performance of the proposed Gaussian Bayes Discriminant Analysis (GBDA) model against existing methods.

Main Methods:

• Developed a Gaussian Bayes Discriminant Analysis (GBDA) model by integrating a Gaussian prior model into Bayes Discriminant Analysis (BDA).
• Applied and compared BDA and GBDA models using high-resolution GF-2 and Sentinel-2 satellite imagery.
• Evaluated extraction accuracy using Overall Accuracy (OA) and Kappa coefficient, comparing with Support Vector Machine (SVM) and Random Forest (RF) methods.

Main Results:

• GBDA achieved the highest accuracy on GF-2 data (OA: 97.84%, Kappa: 0.957) and Sentinel-2 data (OA: 93.51%, Kappa: 0.870).
• The GBDA model significantly outperformed SVM, RF, and BDA in impervious surface extraction accuracy.
• GBDA demonstrated enhanced model robustness, generalization ability, and effectively reduced shadow misclassification.

Conclusions:

• The GBDA model provides a highly reliable method for extracting impervious surfaces from high-resolution remote sensing images.
• This research highlights the practical application value of Bayes discriminant analysis in impervious surface mapping.
• The study offers valuable technical support for obtaining high-quality, high-resolution impervious surface information crucial for urban management.