Point Cloud Ground Segmentation

This ROS 2 module performs ground segmentation on LiDAR point clouds using the ground segmentation algorithm from the Robust Field Autonomy Lab, originally developed as part of the lego_loam SLAM system.

• The segmentation algorithm was proposed by Michael Himmelsbach and the publication is available at https://ieeexplore.ieee.org/document/5548059:
• The vertical structures identification was proposed by Igor Bogoslavskyi and the publication is available at https://ieeexplore.ieee.org/document/7759050:

Overview

Ground segmentation is a critical preprocessing step in many robotics applications such as autonomous navigation and SLAM. This implementation leverages the ground segmentation technique from LeGO-LOAM, which classifies ground and non-ground points in 3D LiDAR scans based on geometric features and elevation gradients.

Features

• Efficient ground/non-ground point classification
• Compatible with standard point cloud formats (e.g., ROS sensor_msgs/PointCloud2)
• Designed for real-time performance in outdoor environments
• Can be integrated as a preprocessing node in SLAM or obstacle detection pipelines

ROS Interface

Subscribed Topics

• /lidar/points (sensor_msgs/PointCloud2)
  The input point cloud data, typically from a 3D LiDAR sensor. This topic should contain the full point cloud from which ground and segmented (non-ground) points will be extracted.

Published Topics

• /ground_cloud (sensor_msgs/PointCloud2)
  Contains only the points classified as ground. Useful for mapping flat surfaces or terrain.

• /segmented_cloud_pure (sensor_msgs/PointCloud2)
  Contains the remaining non-ground points (e.g., obstacles, structures, vegetation).

---

Parameters

laser Namespace

Parameter Name	Type	Default	Description
num_vertical_scans	int	16	Number of vertical scan lines (rings) in the LiDAR sensor (e.g., Velodyne VLP-16).
num_horizontal_scans	int	1800	Number of horizontal resolution bins. Higher values provide finer angular resolution.
vertical_angle_bottom	double	-15.0	Vertical angle (in degrees) of the lowest LiDAR scan ring.
vertical_angle_top	double	15.0	Vertical angle (in degrees) of the highest LiDAR scan ring.
sensor_mount_angle	double	0.0	Mounting angle of the LiDAR sensor (in degrees). Adjust if the sensor is tilted.
ground_scan_index	int	7	Index of the scan ring up to which points are considered for ground segmentation.

image_projection Namespace

Parameter Name	Type	Default	Description
segment_valid_point_num	int	5	Minimum number of consecutive valid points required in a segment to consider it valid.
segment_valid_line_num	int	3	Minimum number of different scan lines that must contain valid points in a segment.
segment_theta	double	60.0	Angle threshold (in degrees) used to determine whether a group of points belongs to the same segment.

---

All parameters can be configured via a YAML configuration file or dynamically on the ROS parameter server.

References

• M. Himmelsbach, F.V. Hundelshausen, and H-J. Wuensche, “Fast Segmentation of 3D Point Clouds for Ground Vehicles,” Proceedings of the IEEE Intelligent Vehicles Symposium, pp. 560-565, 2010.
• I. Bogoslavskyi and C. Stachniss, “Fast Range Image-based Segmentation of Sparse 3D Laser Scans for Online Operation,” Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 163-169, 2016.
• Shan, T. & Englot, B. (2018). LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain. IEEE IROS

License

This project is licensed under the GPLv3 license.