Simultaneous Localization and Mapping in unmodified environments using stereo vision

In this paper we describe an approach that builds three dimensional maps using visual landmarks extractedfrom images of an unmodi?ed environment. We propose a solution to the Simultaneous Localization andMapping (SLAM) problem for autonomous mobile robots using visual landmarks. Our map is representedby a set of three dimensional landmarks referred to a global reference frame, each landmark contains a visualdescriptor that partially differentiates it from others. Signi?cant points extracted from stereo images are usedas natural landmarks, in particular we employ SIFT features found in the environment. We estimate boththe map and the path of the robot using a Rao-Blackwellized particle ?lter, thus the problem is decomposedinto two parts: one estimation over robot paths using a particle ?lter, and N independent estimations overlandmark positions, each one conditioned on the path estimate. We actively track visual landmarks at a localneighbourhood and select only those that are more stable. When a visual feature has been observed from asigni?cant number of frames it is then integrated in the ?lter. By this procedure, the total number of landmarksin the map is reduced, compared to prior approaches. Due to the tracking of each landmark, we obtain differentexamples that represent the same natural landmark. We use this fact to improve data association. Finally,efficient resampling techniques have been applied, which reduces the number of particles needed and avoidsthe particle depletion problem.