Preserve point-level structure
Start from consecutive point clouds without collapsing the scene into objects.
← Back to projects
LiDAR scene flow · Direct point matching
DirectFlowMatch
A deterministic, geometry-aware method that estimates per-point motion from consecutive LiDAR scans without segmentation or clustering.
Sequential LiDARPer-point scene flowGeometry-aware matchingPublic nuScenes validation
00 / Interactive trace
See DirectFlowMatch resolve the scene.
This vehicle-rich nuScenes trace has the strongest ground-truth motion in the curated qualitative set. Begin with two LiDAR scans, inspect the initial point matching, then advance the complete source cloud along the final motion field.
nuScenes · High-motion traffic trace
From two scans to one complete point-level match.
Inspect the input pair, initial matching, and complete match.
Active view 00 · Input pair Two consecutive, voxel-downsampled LiDAR scans.
Preparing the interactive DFM trace…
01 / At a glance
Geometry-derived motion, point by point.
DirectFlowMatch uses the acquisition geometry of consecutive LiDAR scans to organize the search for correspondences, reject implausible candidates, and recover a velocity for each matched point.
InputConsecutive LiDAR scans
OutputPer-point scene flow / velocity
MethodDeterministic direct matching
ValidationPublic nuScenes data
02 / Problem
LiDAR provides geometry, not native Doppler.
A direct time-of-flight LiDAR scan captures the shape of a scene but does not provide the native Doppler measurement. Motion therefore has to be recovered across frames while accounting for sparsity, partial observations, and acquisition geometry.
Constrain correspondence
Use angular neighborhoods and physically meaningful motion bounds.
Recover velocity
Convert matched displacements into per-point motion estimates.
03 / Method
A five-step geometry and consistency pipeline.
Prepare consecutive scans
Bring the two point clouds into a common geometric frame while preserving point-level structure.
Build local correspondences
Search nearby points using acquisition geometry and physically meaningful motion limits.
Check both directions
Compare forward and reverse matches so locally ambiguous or inconsistent motion can be recognized.
Protect coherent motion
Keep evidence-supported moving points from being pulled toward a static-scene explanation.
Recover per-point flow
Select the stable correspondence for each point and convert displacement over time into velocity.
04 / Validation
Evaluated on public nuScenes data.
This dynamic urban trace comes from the public nuScenes dataset. It was selected because the moving foreground makes the improvement from the initial correspondence field to the complete DirectFlowMatch result easy to inspect.
nuScenes · Dynamic urban trace
Inspect the initial and refined point-level motion.
Compare the input scans, the initial matching field, and the complete DirectFlowMatch result. Rotate, zoom, and change point size while reviewing the scene in XYZ.
Active view 00 · Input pair Two consecutive, voxel-downsampled LiDAR scans.
Preparing the interactive DFM trace…
05 / Original platform context
DirectFlowMatch on Valeo’s Scala 2 data.
The algorithm was initially designed for Valeo’s Scala 2 LiDAR data and documented as part of the thesis work. This earlier qualitative view shows consecutive scans aligned for direct point-level correspondence.
06 / Relationship
A LiDAR reference for motion-aware radar work.
The method provides geometry-derived motion for radar supervision and validation, and for downstream tasks that require motion estimates from consecutive dToF LiDAR scans.