Master Thesis opportunity
Master Thesis Opportunities – Gothenburg, Sweden
Classification of geometry as static or dynamic
Classification of surface samples into static or dynamic surfaces simplifies decisionmaking. For static geometry, e.g., the road surface and permanent infrastructure, the classification and response can be pre-computed from measurements in the past, fused from several time steps and from sensors on many vehicles. By classifying the measurements from the lidar sensor as belonging to permanent static geometry, new static geometry, or dynamic geometry, decisions can be made with more information than the sensors of a single vehicle can produce.
The aim is to pre-compute a surface representation of static geometry from an atlas of lidar data. At drive-time, the surface representation can be used to determine the likelihood that an individual lidar measurement belongs to static or dynamic geometry.
The result should be computed in real time within the limits of an in-car computer, or a high-end laptop, and its restrictions on computational resource and memory consumption.
The scope is limited to point cloud from a single lidar sensor mounted on a test vehicle. The sensors relative position and orientation relative to the vehicle are known from calibration, and the vehicles position and orientation in the world are determined for each time step by other methods, so the scope does not include simultaneous localization and mapping.
Decision and control is not within the scope of this thesis.
Real-time 3D Reconstruction in Dynamic Scenes using Point-based Fusion http://reality.cs.ucl.ac.uk/pr...
Further information and contacts
Please send in individual applications. If you wish to partner with someone, simply note that in your application.
For questions regarding the project, please Contact Ognjan Hedberg, firstname.lastname@example.org.