P6. Efficient Representation and Generation of consistent 3D and 4D Maps
Homepage: http://cg.cs.uni-bonn.de/en/projects/mapping-on-demand
Project coordinator/staff: R. Klein, D. Cremers, R. Vock, S. Ochmann
Summary:
The project's goal is to develop and test procedures and algorithms for the fast three-dimensional identification and mensuration of inaccessible objects on the basis of a semantically specified user query. The sensor platform is a lightweight autonomously flying drone. It uses the visual information from cameras for navigation, obstacle detection, exploration and object acquisition.
Our group deals with the efficient representation, management and visualization of 3D surface data that gets captured incrementally. This data will be integrated into a global 3D map. This task can be divided into three parts:
- Developing a compact and efficient data structure for the geometry which was captured and reconstructed during several flights. This data structure shall permit high-performance visualization and range queries. These are necessary to utilize the map in the context of navigation, mission planning, surface reconstruction and scene interpretation. The data structure should provide continuous levels of detail (LOD) of the geometry.
- Detecting and treating inconsistencies in the geometry data which was captured during multiple flights. This includes especially detecting whether observed objects have changed relevantly between flights as well as resolving temporal inconsistencies which can occur when capturing trees and bushes that are moving in the wind.
- Integrating geometry and color data captured during different points in time. Compensating illumination conditions during capturing time to prevent inconsistencies in the reconstructed model's colors as well as permitting the simulation of new illumination conditions.