[Road Extraction in Remote Sensing Images Based on Spectral and Edge Analysis]

Area of Science:

• Remote Sensing
• Computer Vision
• Urban Planning

Context:

• Road extraction from high-resolution imagery is vital for urban planning and transportation.
• Traditional methods struggle with spectral confusion, making road identification difficult.
• Edge features offer crucial information for distinguishing linear objects like roads.

Purpose:

• To develop an improved road extraction method by integrating spectral and edge statistical features.
• To leverage the self-adaptive mean-shift algorithm for robust edge detection.
• To enhance classification accuracy using Support Vector Machine (SVM) with combined features.

Summary:

• A novel method combines spectral information with edge statistical features derived from self-adaptive mean-shift edge detection.
• Edge statistical features capture length and angle distributions, reducing pseudo-edges and noise.
• Support Vector Machine (SVM) classification integrates both feature types for accurate road extraction.

Impact:

• The proposed method achieved an overall accuracy of 93%, a significant improvement over traditional methods (78%).
• Demonstrates the value and efficiency of integrating spectral and edge statistical features for road extraction.
• Particularly effective for road extraction in high-resolution urban images.