Drawing-Based Procedural Modeling of Chinese Architectures.

Area of Science:

• Computer Vision
• Computer Graphics
• Architectural Modeling

Background:

• Traditional methods for 3D architectural modeling often rely on manual input or symbol-based recognition.
• Generating detailed 3D models from 2D elevation drawings, especially for complex architectures like traditional Chinese buildings, remains a challenge.

Purpose of the Study:

• To develop a novel modeling framework for automatically generating 3D models of Chinese architecture from elevation drawings.
• To integrate automatic drawing recognition with procedural modeling techniques for rule extraction and 3D model construction.

Main Methods:

• Utilizing repetitive pattern trees for bottom-up clustering of repetitive regions in elevation drawings to form architectural components.
• Employing a hierarchical approach with shape trees and Markov Random Fields (MRFs) for recognizing global architectural structure and component interdependencies.
• Deriving shape grammar rules to construct 3D semantic models and their variations using a 3D component repository.

Main Results:

• Successfully extracted production rules from elevation drawings through pattern recognition and hierarchical structuring.
• Generated 3D semantic models of Chinese architecture with inherent variations.
• Demonstrated a novel integration of procedural modeling and elevation drawing recognition.

Conclusions:

• The proposed framework offers an effective approach for automated 3D Chinese architectural model generation from elevation drawings.
• The novel integration of repetitive pattern trees, shape trees, and MRFs advances the field of architectural modeling.
• This method provides a foundation for creating diverse 3D semantic models and exploring architectural variations.