Our approach is to merge vision and odometry information using a Kalman-Filter while combining the strengths of both methods, i.e. the high tracking frequency of odometry and the accuracy of vision. Intuitively, the vision information is used to correct the odometry estimations making the resulting local- ization more accurate and yet fast. To evaluate our approach, we use a version of the Festo Robotino with a Guppy PRO F- 125 camera which is mounted on top of the robot (see Figure 1). We compare the localization results of our approach, the Kalman-filtered Vision and Odometry localization (KVO), with the highly accurate DTrack system as ground truth