An autonomous navigation system for a mobile robot is able to build a map and localize in an unknown environment. Through this paper we present our method for mapping and localization of mobile robots using stereo vision. For mapping we used the hybrid card, consisting of nodes and arcs. Construction of a node is composed by extracting SIFT points and calculates their space projections for depth information. Arcs are the connections between the nodes i.e. the distance in the plane between the nodes to each other and they are calculated using odometry. Based on this mapping we used a method based on singular value decomposition for global localization. The results of our experimental test showed the validation of our method on mapping and localization of mobile robots based on stereo vision.