Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GroundGrid:LiDAR Point Cloud Ground Segmentation and Terrain Estimation

Nicolai Steinke, Daniel Göhring, Raùl Rojas

TL;DR

GroundGrid introduces an online, elevation-map based solution for LiDAR ground segmentation and terrain estimation. By iteratively filtering outliers, rasterizing to a 2D grid, and using variance-based ground-cell detection plus a confidence-weighted elevation update with terrain interpolation, it produces accurate ground masks and terrain maps. The approach achieves state-of-the-art IoU on SemanticKITTI (≈94.8%) and RMSE improvements on ALS terrain data (urban ≈0.196 m, hill ≈0.488 m) while running at 171 Hz, enabling real-time perception for autonomous systems. The method is sensor-agnostic, deterministic, and available as open-source, highlighting practical impact for online drivability assessment and obstacle prediction.

Abstract

The precise point cloud ground segmentation is a crucial prerequisite of virtually all perception tasks for LiDAR sensors in autonomous vehicles. Especially the clustering and extraction of objects from a point cloud usually relies on an accurate removal of ground points. The correct estimation of the surrounding terrain is important for aspects of the drivability of a surface, path planning, and obstacle prediction. In this article, we propose our system GroundGrid which relies on 2D elevation maps to solve the terrain estimation and point cloud ground segmentation problems. We evaluate the ground segmentation and terrain estimation performance of GroundGrid and compare it to other state-of-the-art methods using the SemanticKITTI dataset and a novel evaluation method relying on airborne LiDAR scanning. The results show that GroundGrid is capable of outperforming other state-of-the-art systems with an average IoU of 94.78% while maintaining a high run-time performance of 171Hz. The source code is available at https://github.com/dcmlr/groundgrid

GroundGrid:LiDAR Point Cloud Ground Segmentation and Terrain Estimation

TL;DR

GroundGrid introduces an online, elevation-map based solution for LiDAR ground segmentation and terrain estimation. By iteratively filtering outliers, rasterizing to a 2D grid, and using variance-based ground-cell detection plus a confidence-weighted elevation update with terrain interpolation, it produces accurate ground masks and terrain maps. The approach achieves state-of-the-art IoU on SemanticKITTI (≈94.8%) and RMSE improvements on ALS terrain data (urban ≈0.196 m, hill ≈0.488 m) while running at 171 Hz, enabling real-time perception for autonomous systems. The method is sensor-agnostic, deterministic, and available as open-source, highlighting practical impact for online drivability assessment and obstacle prediction.

Abstract

The precise point cloud ground segmentation is a crucial prerequisite of virtually all perception tasks for LiDAR sensors in autonomous vehicles. Especially the clustering and extraction of objects from a point cloud usually relies on an accurate removal of ground points. The correct estimation of the surrounding terrain is important for aspects of the drivability of a surface, path planning, and obstacle prediction. In this article, we propose our system GroundGrid which relies on 2D elevation maps to solve the terrain estimation and point cloud ground segmentation problems. We evaluate the ground segmentation and terrain estimation performance of GroundGrid and compare it to other state-of-the-art methods using the SemanticKITTI dataset and a novel evaluation method relying on airborne LiDAR scanning. The results show that GroundGrid is capable of outperforming other state-of-the-art systems with an average IoU of 94.78% while maintaining a high run-time performance of 171Hz. The source code is available at https://github.com/dcmlr/groundgrid
Paper Structure (15 sections, 11 equations, 4 figures, 2 tables)

This paper contains 15 sections, 11 equations, 4 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. GROUND SEGMENTATION AND TERRAIN ESTIMATION
  4. Outlier Filtering
  5. Point Cloud Rasterization
  6. Classification of Ground Cells
  7. Ground Elevation Calculation
  8. Terrain Interpolation
  9. Point Cloud Segmentation
  10. EXPERIMENTS AND EVALUATION
  11. Parameters
  12. Ground Segmentation Evaluation
  13. Terrain Estimation Evaluation
  14. Run-Time
  15. CONCLUSION

Figures (4)

  • Figure 1: Visualization of outlier points (cyan) below the ground (red) due to a reflection on the car's body (purple). SemanticKITTI seq. 00 cloud 263.
  • Figure 2: Overview of the proposed system. Input point cloud in red, intermediate results in blue, and output segmented point cloud in green.
  • Figure 3: Ground segmentation qualitative evaluation (ground points in green / obstacle points in red)
  • Figure 4: Terrain estimation qualitative evaluation