Efficient and robust drivable region extraction for autonomous vehicles

Nowadays, the autonomous vehicle (AV) is investigated actively by the researchers from all over the world. However, many researches are based on expensive equipments, such as Radar, LIDAR and IMU. It makes the autonomous vehicle's marketization difficult. Thus our paper proposes a low-cost method based on images to obtain the vanishing point of roads and then extract the drivable region. Combining the Hough transform, the approach takes advantage of the image gradient, although from low-resolution images, it can obtain salient features. In addition, the proposed method also can help to detect a single lane line marking. Experiments show the proposed approach can extract drivable region efficiently and stably.