Abstract: Most existing road segmentation algorithms based on vanishing point estimation demand the vanishing point locats inside the image, and they are always time-consuming and cannot effectively overcome the interference of noise which has strong texture features. This paper focuses on these problems, and proposes a road segmentation method based on principal orientation and vanishing point estimation. Firstly, the valid voters are selected by the restrains of road principal orientation. Then a multi-dimension voting scheme is presented, which records the voting information in different orientations of candidate vanishing point, and these information is later used to judge whether the vanishing point is located inside image. Finally, a boundary fitting strategy based on principal orientation is proposed, which extracts the road region according to the data generated on the multi-dimension voting stage. Quantitative and qualitative experiments show that the proposed road segmentation method is more accurate and faster than the traditional algorithms, and it can still work well when the vanishing point is located outside the image.