Tables, Counters, and Shelves: Semantic Mapping of Surfaces in 3D
Trevor, Alexander J. B.
Rogers, John G.
Christensen, Henrik I.
MetadataShow full item record
Semantic mapping aims to create maps that include meaningful features, both to robots and humans. We present an extension to our feature based mapping technique that includes information about the locations of horizontal surfaces such as tables, shelves, or counters in the map. The surfaces are detected in 3D point clouds, the locations of which arc optimized by our SLAM algorithm. The resulting scans of surfaces are then analyzed to segment them into distinct surfaces, which may include measurements of a single surface across multiple scans. Preliminary results are presented in the form of a feature based map augmented with a set of 3D point clouds in a consistent global map frame that represent all detected surfaces within the mapped area.