Research Project P19:
3-D Mosaicing

Research Project Goal

Efficient and comfortable acquisition of large 3D scenes is an important topic for many current and future applications like cultural heritage, web applications and 3DTV and therefore it is a hot research topic. To create realistic virtual realities a modeling process has to be performed, which includes building geometrically and photometrically correct models of complex environments (exteriors and interiors). Today this process is primarily done by hand due to the high complexity of these environments. People wanting to use these models have either to build their own limited models or rely existing models which are themselves inaccurate and lacking feature functionality that high resolution modeling demands. The goal of this research project is to develop a 3-D modeling method for reconstructing building models by using 3-D snapshots taken from differrent locations inside and outside a building.

Research Project Scope

The idea is to use our experimental scanning platform to acquire 3-D snapshots inside a medium sized building. Then build a software to extract salient features from the single models, and for registering the snapshots within a 3-D mosaic. Later, we want to visualize a 3-D model as a realistic walkthrough of a large indoor environment.

Data

Raw geometry data of a single snapshot:



Photorealisic, texture mapped 3-D model:



Experimental scanning platform:

Tasks

• Use 3-D scanning platform to collect snapshots of at least on building
• Develop and implement a method for extracting salient features from geometry (e.g. spin images) or texture (e.g. SIFT)
• Develop and implement a registration method that finds the globally optimized location of each snapshot based on the extracted features
• Visualize the reconstructed model as a walkthrough

Research Project Status

student names here

Point of Contact

Sebastian Thrun

Midterm Report

not yet submitted

Final Report

not yet submitted