Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

10 Citations (Scopus)

Abstract

Deep Learning has been massively used for image classification in recent years. The use of deep learning for ground classification from LIDAR point clouds has also been recently studied. However, point clouds need to be converted into an image in order to use Convolutional Neural Networks (CNNs). In state-of-the-art techniques, this conversion is slow because each point is converted into a separate image. This approach leads to highly redundant computation during conversion and classification. The goal of this study is to design a more efficient data conversion and ground classification. This goal is achieved by first converting the whole point cloud into a single image. The classification is then performed by a Fully Convolutional Network (FCN), a modified version of CNN designed for pixel-wise image classification. The proposed method is significantly faster than state-of-the-art techniques. On the ISPRS Filter Test dataset, it is 78 times faster for conversion and 16 times faster for classification. Our experimental analysis on the same dataset shows that the proposed method results in 5.22% of total error, 4.10% of type I error, and 15.07% of type II error. Compared to the previous CNN-based technique and LAStools software, the proposed method reduces the total error and type I error (while type II error is slightly higher). The method was also tested on a very high point density LIDAR point clouds resulting in 4.02% of total error, 2.15% of type I error and 6.14% of type II error.
Original languageEnglish
Title of host publication2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy
Place of PublicationRiva Del Garda
PublisherInternational Society for Photogrammetry and Remote Sensing (ISPRS)
Pages231-238
Number of pages8
Publication statusPublished - 4 Jun 2018

Publication series

NameISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
PublisherISPRS
VolumeIV-2

Fingerprint

Image classification
Neural networks
Pixels
Deep learning

Keywords

  • ITC-GOLD

Cite this

Rizaldi, A., Persello, C., Gevaert, C. M., & Oude Elberink, S. J. (2018). Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds. In 2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy (pp. 231-238). (ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences; Vol. IV-2). Riva Del Garda: International Society for Photogrammetry and Remote Sensing (ISPRS).
Rizaldi, Aldino ; Persello, C. ; Gevaert, C.M. ; Oude Elberink, S.J. / Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds. 2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy. Riva Del Garda : International Society for Photogrammetry and Remote Sensing (ISPRS), 2018. pp. 231-238 (ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences).
@inbook{c883daa63745404ea6b648f8674f06dc,
title = "Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds",
abstract = "Deep Learning has been massively used for image classification in recent years. The use of deep learning for ground classification from LIDAR point clouds has also been recently studied. However, point clouds need to be converted into an image in order to use Convolutional Neural Networks (CNNs). In state-of-the-art techniques, this conversion is slow because each point is converted into a separate image. This approach leads to highly redundant computation during conversion and classification. The goal of this study is to design a more efficient data conversion and ground classification. This goal is achieved by first converting the whole point cloud into a single image. The classification is then performed by a Fully Convolutional Network (FCN), a modified version of CNN designed for pixel-wise image classification. The proposed method is significantly faster than state-of-the-art techniques. On the ISPRS Filter Test dataset, it is 78 times faster for conversion and 16 times faster for classification. Our experimental analysis on the same dataset shows that the proposed method results in 5.22{\%} of total error, 4.10{\%} of type I error, and 15.07{\%} of type II error. Compared to the previous CNN-based technique and LAStools software, the proposed method reduces the total error and type I error (while type II error is slightly higher). The method was also tested on a very high point density LIDAR point clouds resulting in 4.02{\%} of total error, 2.15{\%} of type I error and 6.14{\%} of type II error.",
keywords = "ITC-GOLD",
author = "Aldino Rizaldi and C. Persello and C.M. Gevaert and {Oude Elberink}, S.J.",
year = "2018",
month = "6",
day = "4",
language = "English",
series = "ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences",
publisher = "International Society for Photogrammetry and Remote Sensing (ISPRS)",
pages = "231--238",
booktitle = "2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy",

}

Rizaldi, A, Persello, C, Gevaert, CM & Oude Elberink, SJ 2018, Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds. in 2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. IV-2, International Society for Photogrammetry and Remote Sensing (ISPRS), Riva Del Garda, pp. 231-238.

Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds. / Rizaldi, Aldino; Persello, C.; Gevaert, C.M.; Oude Elberink, S.J.

2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy. Riva Del Garda : International Society for Photogrammetry and Remote Sensing (ISPRS), 2018. p. 231-238 (ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences; Vol. IV-2).

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

TY - CHAP

T1 - Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds

AU - Rizaldi, Aldino

AU - Persello, C.

AU - Gevaert, C.M.

AU - Oude Elberink, S.J.

PY - 2018/6/4

Y1 - 2018/6/4

N2 - Deep Learning has been massively used for image classification in recent years. The use of deep learning for ground classification from LIDAR point clouds has also been recently studied. However, point clouds need to be converted into an image in order to use Convolutional Neural Networks (CNNs). In state-of-the-art techniques, this conversion is slow because each point is converted into a separate image. This approach leads to highly redundant computation during conversion and classification. The goal of this study is to design a more efficient data conversion and ground classification. This goal is achieved by first converting the whole point cloud into a single image. The classification is then performed by a Fully Convolutional Network (FCN), a modified version of CNN designed for pixel-wise image classification. The proposed method is significantly faster than state-of-the-art techniques. On the ISPRS Filter Test dataset, it is 78 times faster for conversion and 16 times faster for classification. Our experimental analysis on the same dataset shows that the proposed method results in 5.22% of total error, 4.10% of type I error, and 15.07% of type II error. Compared to the previous CNN-based technique and LAStools software, the proposed method reduces the total error and type I error (while type II error is slightly higher). The method was also tested on a very high point density LIDAR point clouds resulting in 4.02% of total error, 2.15% of type I error and 6.14% of type II error.

AB - Deep Learning has been massively used for image classification in recent years. The use of deep learning for ground classification from LIDAR point clouds has also been recently studied. However, point clouds need to be converted into an image in order to use Convolutional Neural Networks (CNNs). In state-of-the-art techniques, this conversion is slow because each point is converted into a separate image. This approach leads to highly redundant computation during conversion and classification. The goal of this study is to design a more efficient data conversion and ground classification. This goal is achieved by first converting the whole point cloud into a single image. The classification is then performed by a Fully Convolutional Network (FCN), a modified version of CNN designed for pixel-wise image classification. The proposed method is significantly faster than state-of-the-art techniques. On the ISPRS Filter Test dataset, it is 78 times faster for conversion and 16 times faster for classification. Our experimental analysis on the same dataset shows that the proposed method results in 5.22% of total error, 4.10% of type I error, and 15.07% of type II error. Compared to the previous CNN-based technique and LAStools software, the proposed method reduces the total error and type I error (while type II error is slightly higher). The method was also tested on a very high point density LIDAR point clouds resulting in 4.02% of total error, 2.15% of type I error and 6.14% of type II error.

KW - ITC-GOLD

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2018/chap/persello_ful.pdf

M3 - Chapter

T3 - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

SP - 231

EP - 238

BT - 2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy

PB - International Society for Photogrammetry and Remote Sensing (ISPRS)

CY - Riva Del Garda

ER -

Rizaldi A, Persello C, Gevaert CM, Oude Elberink SJ. Fully Convolutional Networks for Ground Classification from LiDAR Point Clouds. In 2018 ISPRS TC II Mid-term Symposium “Towards Photogrammetry 2020”, 4–7 June 2018, Riva del Garda, Italy. Riva Del Garda: International Society for Photogrammetry and Remote Sensing (ISPRS). 2018. p. 231-238. (ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences).