The Photogrammetry and Remote Sensing lab implemented an algorithm for the image-based reconstruction of surface meshes. Thereby an initial approximation of the surface is successively refined in a coarse-to-fine scheme. The density of mesh faces is an important parameter with respect to the quality of reconstructions as well as the convergence properties of the algorithm. Within this work, methods for an adaptive simplification of a given surface mesh have to be investigated. The mesh should be simplified in flat areas, for example streets or facades, and densified in areas featuring fine geometric structures. Simplification criteria should involve geometric features of the mesh, semantic information (facade, ground, vegetation, roof) and texture information in the images. The focus is on urban scenes captured by airborne cameras.
The goal of this work is the implementation of algorithms in C/C++ for the adaptive simplification of surface meshes. For this, existing mesh and Computer Vision libraries (OpenMesh, OpneCv, Embree,...) have to be utilized.
Dr. Mathias Rothermel (email@example.com)
CLS Student Project (MPG ETH CLS)
Information, Computing and Communication Sciences
Engineering and Technology