Lane detection based on Improved FCN

Computer vision research has a long history and is still in progress. Lane detection is one of the typical applications of computer vision in the field of traffic, and has a wide range of applications. As one of the important intelligent driving technique, lane detection can not only assist autonomous driving safely, but also give out warning when vehicle is yawing. For lane detection, there are two kinds of major methods, one is based on image processing and the other is based on image segmentation. Because of its good representation and learning ability, the last one has been made great progress in lane detection field recently. Based on the analysis of fully convolutional network (FCN) and conditional random field (CRF), a novel lane detection algorithm is proposed in this paper, which is based on the combination of FCN and CRF. Some extensive experimental results show that the proposed method is robust against shadow, occlusion and slope variations in the public lane dataset.