Robust local mapping using stereo vision

In this paper we present a new method for mapping and localization of mobile robots. Modelling the environment in this work is based on visual information from a stereo sensor. The work is to extract the two images produced by the sensor stereo invariant local descriptors. Using the parameters of both cameras, as determined by calibration, we calculated the matching points between the left and right images and coordinated 3D forming a cloud of 3D points. The latter is the local map of the area of navigation in this position. This map is characterized by the SIFT point that characterizes this area and provides a coordinated 3D geometric information of this space.