Skip to main content

Module ground_segmentation

Module ground_segmentation 

Source
Expand description

Ground segmentation for outdoor LiDAR point clouds.

Implements a Patchwork++-style algorithm that splits a point cloud into ground and non-ground subsets. The approach uses a Concentric Zone Model (CZM): the XY plane around the sensor is divided into concentric rings and angular sectors, producing a grid of patches whose size adapts with range. Each patch is fit with a Region-wise Ground Plane Fit (R-GPF) — seed points near the patch minimum are selected, a plane is fit by PCA, then inliers are extracted and the plane is refit iteratively. Each candidate plane is then validated by three criteria: uprightness (normal nearly vertical), elevation (within an allowed height band per zone), and flatness (smallest eigenvalue ratio).

Reference: Lee et al., “Patchwork++: Fast and Robust Ground Segmentation Solving Partial Under-Segmentation Using 3D Point Cloud”, IROS 2022.

Structs§

GroundSegmentationResult
Result of ground segmentation.
PatchworkConfig
Configuration for the Patchwork++ ground segmentation algorithm.

Functions§

patchwork_plus_plus
Run Patchwork++-style ground segmentation on a point cloud.
segment_ground
Convenience wrapper using default configuration with a given sensor height.