Description
Terrestrial Laserscanning (TLS) is a means of measuring large structures like buildings or archaeological excavation sites. The result of a laserscan is a point cloud, that is a set of 3D coordinates representing the surfaces of scanned objects. When a camera is used, photographs of the scanned objects can be taken, and the photos can be projected onto the scanned points, which results in a colorized point cloud. In the picture above you see a part of the Domitilla Catacomb in Rome, consisting of about 200 million points and 170 scan positions. Rendering such huge models makes use of out-of-core techniques, a specialized data structure, and simple vertex shaders. The point cloud as such holds no information about the geometry of the objects it contains, it is just a set of points.
Task
| The segmentation of a point cloud is a subdivision of the point set into parts that are coherent by some means. Segmentation of an arbitary object is still an open research area, and current approaches are only capable of handling special cases. Building on this work we would like to develop our own segmentation algorithms, since we are using otherwise unprocessed point clouds. You have the chance to work in a creative environment and we are looking forward to the contribution of your ideas. |
Requirements
Good knowledge and high interest in computer graphics and data structures, good programming skills (C++)Additional Information
Contact: Claus ScheiblauerLiterature:
