Many people are involved in severe traffic related accident every year. Many of these accidents are strongly linked to too high speed. One way to decrease the speed transgressions and the number of accidents could be to implement a system that help the driver keeping the current speed limit. This thesis presents a method for classifying speed signs using computer vision. The classifier is a part of a system for speed tracking and notifying. This system assists a driver by notifying him/her of the current speed limit. This is done by computer vision, i.e. a camera connected to a computer reads the current trac environment and the frames are processed by the computer. A computer vision system is commonly composed by a detector and a classifier. The detector scans the frames from the camera and locates possible speed signs. The possible speed sign locations are passed to the classifier. The classifiers task is to decide if the possible location contains a speed sign and if it does contain a speed sign, to proclaim its speed. The classifier in this thesis is based on cross correlation. The algorithm is evaluated against a database, which includes images with coordinates for speed signs marked by hand. If the weather conditions are not to extreme and if the detector has good accuracy the classifier will make the right decision in close to 100% of the cases.