Process Data
Generate Sensor Views
If a trajectory is available, the sensor views (e.g., LiDAR beams) can be reconstructed:
Linux / macOS
cargo run -r -- generate-sensor-views \
--reflection-uncertainty-line-length 1.0The --reflection-uncertainty-line-length sets the length (in meters) of the uncertainty line segment for each sensor beam reflection.
Both generate-sensor-views and associate are connection-intensive operations. If they are slow, increase --db-max-connections (or SENSORDB_MAX_CONNECTIONS) beyond the default of 10 — values around 50 are typical. See the FAQ for tuning guidance.
Associate Observations with Object Surfaces
To associate sensor observations with objects from the semantic model:
Linux / macOS
cargo run -r -- associate \
--reflection-uncertainty-point-buffer 0.5 \
--reflection-uncertainty-line-buffer 0.1 \
--max-reflection-uncertainty-line-intersection-parameter 1.0Parameters
--reflection-uncertainty-line-buffer controls how beams are associated with surfaces:
0.0— only beams that directly intersect the surface geometry are included.> 0— beams are also included when a surface falls within the buffer volume around the reflection uncertainty line.
--max-reflection-uncertainty-line-intersection-parameter prevents associations when beams are nearly parallel to surfaces within the buffer volume.
When a beam is nearly parallel to a surface, the intersection parameter can become extremely large, as the intersection point is projected far beyond the intended beam range. Setting this parameter (e.g., 1.0) caps the maximum acceptable intersection distance, filtering out unrealistic associations caused by grazing-angle beams or inaccuracies.
The intersection parameter represents the distance along the reflection uncertainty line where an intersection occurs with the plane of the surface. The value can also be negative if the reflected point lies behind the surface to be associated.