Detail publikace

CNN for Very Fast Ground Segmentation in Velodyne LiDAR Data

VEĽAS, M. ŠPANĚL, M. HRADIŠ, M. HEROUT, A.

Originální název

CNN for Very Fast Ground Segmentation in Velodyne LiDAR Data

Anglický název

CNN for Very Fast Ground Segmentation in Velodyne LiDAR Data

Jazyk

en

Originální abstrakt

This paper presents a novel method for ground segmentation in Velodyne point clouds. We propose an encoding of sparse 3D data from the Velodyne sensor suitable for training a convolutional neural network (CNN). This general purpose approach is used for segmentation of the sparse point cloud into ground and non-ground points. The LiDAR data are represented as a multi-channel 2D signal where the horizontal axis corresponds to the rotation angle and the vertical axis represents channels - laser beams. Multiple topologies of relatively shallow CNNs (i.e. 3-5 convolutional layers) are trained and evaluated, using a manually annotated dataset we prepared. The results show significant improvement of performance over the state-of-the-art method by Zhang et al. in terms of speed and also minor improvements in terms of accuracy.

Anglický abstrakt

This paper presents a novel method for ground segmentation in Velodyne point clouds. We propose an encoding of sparse 3D data from the Velodyne sensor suitable for training a convolutional neural network (CNN). This general purpose approach is used for segmentation of the sparse point cloud into ground and non-ground points. The LiDAR data are represented as a multi-channel 2D signal where the horizontal axis corresponds to the rotation angle and the vertical axis represents channels - laser beams. Multiple topologies of relatively shallow CNNs (i.e. 3-5 convolutional layers) are trained and evaluated, using a manually annotated dataset we prepared. The results show significant improvement of performance over the state-of-the-art method by Zhang et al. in terms of speed and also minor improvements in terms of accuracy.

Dokumenty

BibTex


@inproceedings{BUT157178,
  author="Martin {Veľas} and Michal {Španěl} and Michal {Hradiš} and Adam {Herout}",
  title="CNN for Very Fast Ground Segmentation in Velodyne LiDAR Data",
  annote="This paper presents a novel method for ground segmentation in Velodyne point
clouds. We propose an encoding of sparse 3D data from the Velodyne sensor
suitable for training a convolutional neural network (CNN). This general purpose
approach is used for segmentation of the sparse point cloud into ground and
non-ground points. The LiDAR data are represented as a multi-channel 2D signal
where the horizontal axis corresponds to the rotation angle and the vertical axis
represents channels - laser beams. Multiple topologies of relatively shallow CNNs
(i.e. 3-5 convolutional layers) are trained and evaluated, using a manually
annotated dataset we prepared. The results show significant improvement of
performance over the state-of-the-art method by Zhang et al. in terms of speed
and also minor improvements in terms of accuracy.",
  address="Institute of Electrical and Electronics Engineers",
  booktitle="IEEE International Conference on Autonomous Robot Systems and Competitions",
  chapter="157178",
  doi="10.1109/ICARSC.2018.8374167",
  edition="NEUVEDEN",
  howpublished="online",
  institution="Institute of Electrical and Electronics Engineers",
  year="2018",
  month="april",
  pages="97--103",
  publisher="Institute of Electrical and Electronics Engineers",
  type="conference paper"
}