tile_ground_segmentation

A ROS1 Noetic package for tile-based ground segmentation of point clouds.

Features

• Tile/grid-based ground estimation for 3D point clouds
• All key parameters are configurable via launch file
• Publishes:
  • /ground_points: segmented ground points
  • /obstacle_points: segmented obstacle points (height above ground > threshold)
  • /ground_est_marker_array: visualization of estimated ground as thin tiles (MarkerArray)
• Robust to missing data and outliers, with slope and neighbor constraints

Main Parameters (see launch/tile_ground_segmentation.launch)

• input_topic: Input point cloud topic
• leaf_size: VoxelGrid downsample size
• roi_x_min, roi_x_max, roi_y_min, roi_y_max, roi_z_min, roi_z_max: ROI limits
• tile_size: Grid cell size (meters)
• z_bin_size: Z bin size for ground detection
• z_bin_min_points: Minimum points in a bin to be considered ground
• max_slope: Maximum allowed ground slope (ratio)
• publish_ground_est_cloud: Whether to publish ground estimation visualization
• obstacle_height_thresh: Height above ground to be considered obstacle (meters)

Usage

1. Build the package with catkin_make.
2. Source your workspace: source devel/setup.bash
3. Launch:

   roslaunch tile_ground_segmentation tile_ground_segmentation.launch

4. Visualize in RViz:
   • Add /ground_points, /obstacle_points as PointCloud2
   • Add /ground_est_marker_array as MarkerArray

Notes

• The algorithm is robust to missing or occluded ground by using neighbor interpolation.
• All logic is implemented in src/tile_ground_segmentation_node.cpp.
• For best results, tune parameters for your sensor and environment.