Vanishing Point Detection Using Angle-based Hough Transform and RANSAC

The information provided by vanishing points is essential for Intelligence Transportation Systems (ITS) such as self-driving or driving scene technology. Vanishing points are representations of infinity points or lines that offer orientation data. Geometrically, a line's vanishing point is found by crossing through the camera's center and intersecting the line of the picture plane. In other words, the vanishing point can be represented by the intersection of two straight lines. Detecting lines in the road image is complex work because the shape of the line is not always straightforward and has much noise. Moreover, vehicles that appear in the image can bother detection. Therefore, this study aims to detect vanishing points using a geometrical-based approach using Hough Transform and Random Sampling Consensus (RANSAC) algorithm and analyze several road conditions that can affect the detection of vanishing points. The lines are restricted at a certain angle. The result shows that RMSE reached 13.33 of 25 images with different traffic conditions. It is a fair result but can be improved. The authors suggest that future study use motion vectors or features of optical flow generate vanishing points better.