Journal of Geodesy and Geoinformation Science ›› 2021, Vol. 4 ›› Issue (3): 49-59.doi: 10.11947/j.JGGS.2021.0305

• Special Issue • Previous Articles     Next Articles

A Simple Deep Learning Network for Classification of 3D Mobile LiDAR Point Clouds

Yanjun WANG1,2,3(),Shaochun LI1,2,3,Mengjie WANG1,2,3,Yunhao LIN1,2,3   

  1. 1. Hunan Provincial Key Laboratory of Geo-Information Engineering in Surveying, Mapping and Remote Sensing, Hunan University of Science and Technology, Xiangtan 411201, China
    2. National-local Joint Engineering Laboratory of Geo-spatial Information Technology, Hunan University of Science and Technology, Xiangtan 411201, China
    3. School of Resource Environment and Safety Engineering, Hunan University of Science and Technology, Xiangtan 411201, China
  • Received:2020-12-14 Accepted:2021-01-15 Online:2021-09-20 Published:2021-10-09
  • About author:Yanjun WANG (1984—), male, PhD, associate professor, majors in LiDAR remote sensing and environmental modeling application. E-mail:
  • Supported by:
    National Natural Science Foundation of China(41971423);National Natural Science Foundation of China(31972951);National Natural Science Foundation of China(41771462);Hunan Provincial Natural Science Foundation of China(2020JJ3020);Science and Technology Planning Project of Hunan Province(2019RS2043);Science and Technology Planning Project of Hunan Province(2019GK2132);Outstanding Youth Project of Education Department of Hunan Province(18B224)


Automatic and accurate classification is a fundamental problem to the analysis and modeling of LiDAR (Light Detection and Ranging) data. Recently, convolutional neural network (ConvNet or CNN) has achieved remarkable performance in image recognition and computer vision. While significant efforts have also been made to develop various deep networks for satellite image scene classification, it still needs to further investigate suitable deep learning network frameworks for 3D dense mobile laser scanning (MLS) data. In this paper, we present a simple deep CNN for multiple object classification based on multi-scale context representation. For the pointwise classification, we first extracted the neighboring points within spatial context and transformed them into a three-channel image for each point. Then, the classification task can be treated as the image recognition using CNN. The proposed CNN architecture adopted common convolution, maximum pooling and rectified linear unit (ReLU) layers, which combined multiple deeper network layers. After being trained and tested on approximately seven million labeled MLS points, the deep CNN model can classify accurately into nine classes. Comparing with the widely used ResNet algorithm, this model performs better precision and recall rates, and less processing time, which indicated the significant potential of deep-learning-based methods in MLS data classification.

Key words: deep learning; convolutional neural network (CNN); mobile laser scanning (MLS); LiDAR data classification; point-to-image transformation