Image-based modeling of unwrappable façades

Area of Science:

• Computer Vision
• 3D Reconstruction
• Geometric Modeling

Background:

• Accurate 3D modeling of building façades is crucial for various applications, including urban planning and digital heritage.
• Existing methods often struggle with non-planar or complex façade geometries, limiting their applicability.

Purpose of the Study:

• To propose a novel unwrappable representation for image-based façade modeling.
• To extend state-of-the-art planar façade modeling techniques to handle more general, non-planar cases.
• To enable detailed reconstruction of complex building façades from registered images.

Main Methods:

• Reconstruction of semidense 3D points from multiple registered images.
• Extraction of mutually orthogonal baseline and profile to define an unwrappable space.
• Mapping 3D points and textures to the unwrapped space for planar modeling.
• Wrapping augmented details back to the original 3D space for final model generation.

Main Results:

• Development of an unwrappable representation that parameterizes façades using baseline and profile.
• Successful reconstruction of complex, non-planar façades by leveraging planar modeling techniques in an unwrapped space.
• Generation of detailed 3D façade models with composed textures from input images.

Conclusions:

• The proposed unwrappable representation effectively extends 3D façade modeling capabilities to non-planar structures.
• This method facilitates detailed reconstruction by temporarily transforming complex geometry into a planar domain.
• The approach offers a powerful new tool for creating accurate and comprehensive 3D building models from image data.