Real-Time Vanishing Point Detector Combining RANSAC and Hough Transform

As a vanishing point(VP)has 2 degrees of freedom(DOF),a real-time VP detection method based on the combination of random sample consensus(RANSAC)and Hough transform is proposed.Firstly,RANSAC is used to extract a set of candidate VP inliers from the edge segments of the image,with each candidate VP inlier fixing one DOF of a VP as the VP is on the extension line of the candidate;Then the Hough transform is conducted along the extension line of each candidate VP inlier to extract the VP’s remaining DOF.The experiment on York Urban Dataset shows that the proposed algorithm is at least 0.69percent better than the state of the art in horizon line accuracy and achieves the processing speed that is at least 90 times as fast as the state of the art.