Three-dimensional outdoor SLAM Using rotation invariant descriptors of salient regions

This research proposes a novel approach to outdoor simultaneous localization and mapping (SLAM) based on local 3-D map matching. An iterative closest point (ICP) algorithm is used to match local 3-D maps and estimate a robot pose, but an alignment error is generated by the ICP algorithm due to the false selection of corresponding points. This paper proposes a new method to extract 3-D points that are valid for ICP matching. Rotation invariant descriptors are introduced for robust correspondence. 3-D environmental data acquired by tilting a 2-D laser scanner are used to build local 3-D maps. Experimental results in real environments show the increased accuracy of the ICP-based matching and a reduction in matching time.