In this paper, we propose a robust curved lane marking detection method by first detecting a straight lane and applying a geometric model of that detected straight lane. In our proposed method, we first detect the straight line and generate 13 candidates of the curved lane by applying a geometric model. We then vote those candidates on the feature image and consider the candidate which acquires the highest number of votes as the final curved lane. The performance and the efficiency of the proposed algorithm are confirmed by the experimental results.