All models Cross-cutting preprocessing
Noise filtering — A clean cloud before any classification
Isolated and building-edge noise detected and removed ahead of every downstream model.
What it does
Every survey carries noise — atmospheric returns, birds, multipath and scan-edge artefacts that corrupt surfaces and inflate volumes. The noise model detects and isolates these returns first, lifting the accuracy of every other model that runs after it. It runs automatically ahead of aerial, mobile and corridor classification.
Available categories
What it classifies.
Every point is assigned a typed class — then vectorised into objects your workflow consumes.
07Low isolated noise (below ground)
18High isolated noise (dust, birds, cloud)
—High building noise (interiors / edges)
23Points below ground
Noise catalogue
Noise patterns, across sensors
The same model isolates low and high noise consistently, whatever the sensor or survey. Detected once, up front, so every downstream model runs on a clean cloud.
Low isolated noisebelow ground
High isolated noisedust, birds, cloud
Unclassifiedvalid returns kept