Our work is dealing with lane markings detection and the vehicle location. We will show how computer vision can improve the accuracy of the determination of the vehicle position in a map by GPS and proprioceptive sensors. An efficient method for locating vehicle by cameras, proprioceptive sensors and GPS has been developed and demonstrated in an outdoor experimental track in real time. The system is designed to a well structured road with lane markings. It merges proprioceptive measurement, GPS location and images analysis information with use of a non linear dynamic model(Kalman Filter). The performance of the system is shown in the experimental track with a processing frequency of 15 Hertz and the error of the location is ±5cm.