This study introduces a novel detection algorithm to recognize the lane markers on the highway at night. The proposed algorithm utilizes adapted median filtering and edged detection of the four direction template logarithmic Prewitt based on the difference of luminous density. It gives the threshold of binaryzation. Determine the ROI and sides lane line, so greatly reduce the amount of computation. We use lane marking width and temporal trajectory strategy to remove false lanes caused by lighting change, and vehicle occlusions. Finally we get complete clear lane line. The proposed method is proved to be fast, stability and accuracy.