3 Radu Bogdan Rusu Andreas Holzbach Michael Beetz Gary Bradski 2009 ICCV, S3DV Workshop perception computer vision robotics manipulation <p>This paper proposes a novel 3D scene interpretation approach for robots in<br />mobile manipulation scenarios using a set of 3D point features (Fast Point Feature<br />Histograms) and probabilistic graphical methods (Conditional Random Fields). <br />Our system uses real time stereo with textured light to obtain dense depth maps<br />in the robot's manipulators working space.  For the purposes of manipulation, we want to <br />interpret the planar supporting surfaces of the scene, recognize and <br />segment the object classes into their primitive parts in 6 degrees of freedom (6DOF) so that <br />the robot knows what it is attempting to use and where it may be handled.<br />The scene interpretation algorithm uses a two-layer classification scheme: i)~we<br />estimate Fast Point Feature Histograms (FPFH) as local 3D point<br />features to segment the objects of interest into geometric primitives; and<br />ii)~we learn and categorize object classes using a novel Global Fast Point Feature<br />Histogram (GFPFH) scheme which uses the previously estimated primitives at each<br />point. To show the validity of our approach, we analyze the proposed system for the<br />problem of recognizing the object class of 20 objects in 500 table settings<br />scenarios.  Our algorithm identifies the planar surfaces, decomposes the scene<br />and objects into geometric primitives with 98.27% accuracy and uses the<br />geometric primitives to identify the object's class with an accuracy of 96.69%.</p>