To solve the problem of lane detection in intelligent auxiliary driving system and as the common ways have some shortcoming such as less real-time and adaptive, the paper raises a new way. Based on the edge being extracted, the feature model of lane markings is built by combining the Median of the Intercepts and the spline. One effective and real time method in highway surface searching and following is presented. The result proves that the method can figure out the way stretching correctly. And therefore, it provides reliable information for the vehicle to be self-anti-deviation. At the same time, it also supplies some valuable discussion on problems existing in its experiments.