Abstarct: Vehicle speed estimation, plays an important role in traffic control and law enforcement. It is a hot but complex topic in intelligent transportation systems. One solution for detecting speed of vehicles is installing cameras on highways, finding vehicles and tracing them in a period of time using image processing techniques. The advantages of these techniques are saving money and less use of operators to monitor real-time video images. In this thesis, we want to estimate speed of vehicles with extraction of vanishing points in the image. It is used to calculate internal camera parameters and consequently speed of vehicles entering to the region of interest (ROI). The mean error of our algorithm is about 10%. In addition, we also provide a real time method for counting on-roads vehicles.