Abstract : The autonomous navigation ability and road following precision require two skills, the obstacle avoidance skill and localization and map following skill. In this thesis a mobile robot navigation system is designed. The robot is able to navigate safely and successfully in an unstructured environment relying only on uncalibrated vision camera as its only mean of perception. For obstacle avoidance, the proposed algorithm makes use of the strong points in the floor plane extraction algorithm in detecting obstacles. For avoiding obstacles, the strong points of the vector field histogram algorithm along with the fuzzy logic algorithm are considered. A landmark recognition routine is implemented for localization and map following and each landmark is considered a node in the topological map that is fed to the system. A general navigation routine is implemented in order to improve the navigation operation. This routine sets priorities for different situation to eliminate the conflict that may occur between the obstacle avoidance and the map following routines. Practical experiments have been conducted using the manufactured mobile platform (MARTIN). MARTIN was able to navigate successfully in indoor environment. The floor plane extraction algorithm is very efficient in flat ground. However there are certain drawbacks due to the presence of obstacles that cannot be detected using vision only. The results illustrate the validity of the presented framework.