Weighted Iterative CD-Spline for Mitigating Occlusion Effects on Building Boundary Regularization Using Airborne...

Area of Science:

• Computer Vision
• Geographic Information Systems (GIS)
• Remote Sensing

Background:

• Building boundary regularization accuracy is often reduced by occlusions.
• Existing methods struggle to effectively mitigate the impact of occlusions on boundary modeling.
• Accurate building boundary extraction is crucial for various geospatial applications.

Purpose of the Study:

• To propose a novel Weighted Iterative Changeable Degree Spline (WICDS) approach for accurate building boundary modeling.
• To address and minimize errors introduced by occlusions in building contour extraction.
• To evaluate the performance of WICDS using simulated and real-world data.

Main Methods:

• Developed a WICDS approach utilizing a weight function for initial building boundary points.
• Assigned lower weights to points within occlusion regions to reduce their influence.
• Conducted experiments with both simulated and real geospatial data.

Main Results:

• The WICDS approach demonstrated significant potential in modeling building boundaries with occlusions, including curved segments.
• Achieved high performance metrics, with F-score around 99% and PoLiS metric around 0.19 m.
• Showed improvements over the previous Iterative Changeable Degree Spline (ICDS) method, including a 6.5% increase in completeness and 4% in F-score.

Conclusions:

• The WICDS method offers a robust solution for accurate building boundary modeling in the presence of occlusions.
• The proposed approach enhances contour modeling accuracy and outperforms previous methods.
• WICDS shows promise for applications requiring precise building footprint extraction from imagery.