This content is not included in your SAE MOBILUS subscription, or you are not logged in.

LiDAR Data Segmentation in Off-Road Environment Using Convolutional Neural Networks (CNN)

Journal Article
2020-01-0696
ISSN: 2641-9637, e-ISSN: 2641-9645
Published April 14, 2020 by SAE International in United States
LiDAR Data Segmentation in Off-Road Environment Using Convolutional Neural Networks (CNN)
Citation: Dabbiru, L., Goodin, C., Scherrer, N., and Carruth, D., "LiDAR Data Segmentation in Off-Road Environment Using Convolutional Neural Networks (CNN)," SAE Int. J. Adv. & Curr. Prac. in Mobility 2(6):3288-3292, 2020, https://doi.org/10.4271/2020-01-0696.
Language: English

Abstract:

Recent developments in the area of autonomous vehicle navigation have emphasized algorithm development for the characterization of LiDAR 3D point-cloud data. The LiDAR sensor data provides a detailed understanding of the environment surrounding the vehicle for safe navigation. However, LiDAR point cloud datasets need point-level labels which require a significant amount of annotation effort. We present a framework which generates simulated labeled point cloud data. The simulated LiDAR data was generated by a physics-based platform, the Mississippi State University Autonomous Vehicle Simulator (MAVS). In this work, we use the simulation framework and labeled LiDAR data to develop and test algorithms for autonomous ground vehicle off-road navigation. The MAVS framework generates 3D point clouds for off-road environments that include trails and trees. The important first step in off-road autonomous navigation is the accurate segmentation of 3D point cloud data to identify the potential obstacles in the vehicle path. We use simulated LiDAR data to segment and detect obstacles using convolutional neural networks (CNN). Our analysis is based on SqueezeSeg, a CNN-based model for point cloud segmentation. The CNN was trained with a labelled dataset of off-road imagery generated by MAVS and evaluated on the simulated dataset. The segmentation of the LiDAR data is done by point-wise classification and the results show excellent accuracy in identifying different objects and obstacles in the vehicle path. In this paper, we evaluated the segmentation performance at different LiDAR vertical resolutions: the 8-beam and 16-beam. The results showed that there is about 5% increase in accuracy with 16-beam sensors compared with the 8-beam sensors in detecting obstacles and trees. However, the 8-beam LiDAR performance is comparable with the 16-beam sensor in segmenting vegetation, trail-road and ground.