In order to improve the speed of vehicles locating at intersection, this paper proposes a method based on two-dimensional projection vehicle model (TPVM) and optical flow. Firstly, the transformation model between road coordinate system and pixel coordinate system is established on the basis of the internal and external camera parameters which got by off-line calibration. Secondly, the TPVM in the road plane is founded according to the corner points of the vehicle, with the discarding of height information of the vehicle. The coordinates of the model’s corner points in the road plane are determined by the former transformation model, according to the pixel coordinates of the corner points belong to this model. Furthermore, to determine its coordinates in the road coordinate system at any time, the TPVM was tracked by optical flow. Finally, the experiment is designed to verity the locating accuracy and speed of this method. This experiment compared the two factors of TPVM method proposed this paper with those of two existing methods. The result shows that the errors of the three methods are 60.45 cm, 74.24 cm and 66.34 cm respectively. Meanwhile, the locating speed of TPVM method increase 20.33 frames and 12.47 frames compared to the two other methods，separately. The experiment proves that the method can improve the speed of vehicle locating at the intersection in the premise of ensuring the locating accuracy.

CICTP