Performance comparison of point and plane features for SLAM /

Yörük, Mücahit
Simultaneous Localization and Mapping (SLAM) is an indispensable capability for mobile robots that explore unknown environments. This advanced method is now widely employed since the development of improvements in sensor technology, such as 3D depth cameras. To avoid the risk of the human interaction in dangerous environments, various SLAM algorithms have been developed and proposed in the literature. The aim of this study, is to develop a landmark vector that improves the SLAM performance using the planar features of objects. In order to achieve this goal we generated a fastSLAM algorithm and two different feature extraction methods. The first feature extraction method is SURF, which gives responses at the edges of the depth images and the second feature extraction method is plane detection, which gives a compact representation of the environment. Throughout this thesis, four different landmark vectors are defined (SURF point, plane as point, plane as oriented point and plane as surface) and compared the effects on the SLAM. The advantages of using planar features are shown with both the RGBD SLAM dataset and the real time application.
Citation Formats
M. Yörük, “Performance comparison of point and plane features for SLAM /,” M.S. - Master of Science, Middle East Technical University, 2014.