Photorealistic large-scale urban city model reconstruction

Area of Science:

• Computer Graphics
• Virtual Reality
• Geographic Information Systems

Background:

• Realistic virtual environments are essential for civil and defense applications like training and simulations.
• Current methods for creating large-scale virtual environments are time-consuming and manual.
• Accurate environmental representation enhances user immersion and bridges the gap between physical and virtual realities.

Purpose of the Study:

• To propose a novel method for the rapid reconstruction of photorealistic large-scale virtual environments.
• To automate the identification and reconstruction of building structures.
• To enhance texture realism by recovering missing or occluded information.

Main Methods:

• Development of an extendible, parameterized geometric primitive for automatic building identification and reconstruction.
• Interactive reconstruction of complex building roofs using linear and nonlinear primitives.
• A rendering pipeline that integrates data from ground, aerial, and satellite sensors to recover texture information.

Main Results:

• Successful rapid reconstruction of photorealistic large-scale virtual environments.
• Automatic identification and reconstruction of building structures, including complex roofs.
• Enhanced texture quality through the integration of multi-sensor data, effectively handling occlusions.

Conclusions:

• The proposed method significantly accelerates the creation of detailed virtual environments.
• The approach offers a robust solution for reconstructing complex architectural features and textures.
• This technique has broad applicability in various virtual reality domains, improving realism and efficiency.