Heatmap-based Vanishing Point boosts Lane Detection

  • 2020-07-30 17:17:00
  • Yin-Bo Liu, Ming Zeng, Qing-Hao Meng
  • 8


Vision-based lane detection (LD) is a key part of autonomous drivingtechnology, and it is also a challenging problem. As one of the importantconstraints of scene composition, vanishing point (VP) may provide a usefulclue for lane detection. In this paper, we proposed a new multi-task fusionnetwork architecture for high-precision lane detection. Firstly, the ERFNet wasused as the backbone to extract the hierarchical features of the road image.Then, the lanes were detected using image segmentation. Finally, combining theoutput of lane detection and the hierarchical features extracted by thebackbone, the lane VP was predicted using heatmap regression. The proposedfusion strategy was tested using the public CULane dataset. The experimentalresults suggest that the lane detection accuracy of our method outperformsthose of state-of-the-art (SOTA) methods.


Quick Read (beta)

loading the full paper ...