Multi-lane recognition system gives a significant assistance for driver’s safety. In this paper, we try to use a dual parabolic model for more precious and robust lane description. For gathering an efficient feature, we used a parallel RANSAC algorithm. The line model characterization is combined with adjacent point histogram peak analysis and two-axis parabola. Proposed algorithm can detect lane automatically and adaptively in every frame. The designed method is applied using various video images captured from the commercial black box camera and is verified to be robust.