Logo Geosciences


Logo Rennes1
Logo Doc OSUR

Géosciences Rennes
UMR 6118
Université de Rennes1
Campus de Beaulieu
35042 Rennes Cedex

02 23 23 60 76


Sur ce site

Sur le Web du CNRS

Accueil du site > Français > Annuaire > LAGUE Dimitri > 3D Point Cloud Classification with CANUPO

3D Point Cloud Classification with CANUPO

The Canupo software page, user guide and classifiers


Well, the above picture says all : take a 3D point cloud, and classify its points according to a geometrical property called multiscale dimensionality. An important application is for vegetation detection in large 3D point clouds (TLS & ALS), but it is more generic than that. For more information read :

CANUPO exists in 2 flavors :

The two versions are compatible : you can use a classifier description file (.prm) created in CANUPO with qCANUPO, and vice-versa.

Video Tutorials

I’m working on it !

Existing classifiers

The software use parameter files constructed by different users to separate two types of features in a point cloud scene. Here is a list of classifiers that I have constructed for my own application. You are free to use them for non-commercial applications. For commercial applications, contact me.

CHECK the application AND SCAN RESOLUTION. These examples may work for you. if not you’ll have to create a new classifier for your application (see tutorial video)

Classifiers for HIGH-RESOLUTION Terrestrial Lidar

In construction Download latest version of TLS_CLASSIFIERS

- Generic grass/shrub vegetation classifiers (Veget_Tidal.prm) : developped in tidal environment for centimeter resolution scans. Scale range of 5-50 cm. Close to the vegetation classifier for tidal shrubs in Brodu and Lague, 2012, but has been improved to work also on channel banks. Ex : Grass on river bank

- Generic large vegetation classifier (Veget_RangiCliff.prm) : developped in mountain environments for centimeter resolution scans to detect bushes and trees. Scale range of 0.05-1.5 m. Used in the Rangitikei gorge example in Lague, Brodu and Leroux, in press. As it uses a very large maximum scale (for tree detection), it could be slow and memory consumming on very large high resolution clouds. ex : Trees and bushes on a cliff

- Generic large vegetation classifier for long-range scanners (Veget_Longrange.prm) : developed for vegetation on cliffs (bushes, trees) scanned at decimeter resolution. Scale range : 0.2-1.5 m. Preliminary version, still has some trouble with trunks. Worth trying the Generic large vegetation classifier too. Ex : Trees and bushed on Complex 3D cliff

- Otira Gorge Classifiers as in Brodu and Lague, 2012, developped for centimeter resolution scans. Scale range : 2-100 cm. See top picture.

  • Vegetation classifier (veget_super.prm) : shrubs, grass, bushes
  • Bedrock classifier : to separate flat surface from rough surfaces (gravel, water)
  • Gravel/water classifier : to separate gravel from water


These work on high resolution lidar (i.e., at least 10 pts/m²). They are particularly suited to complex environments (mountains, steep slopes,...) where traditional vegetation classification softwares (e.g. Terrascan) may not work really well. Use them at your own risk

In construction Download latest version of ALS_CLASSIFIERS

User guide and tutorial for the command line version

Old stuff but worth checking if you want to improve your classifiers. Version 1.2 of the user guide and tutorial.

PDF - 1.6 Mo
Classification tutorial, v1.2, april 2013

Zip file containing the directory structure and batch files as described in the user guide. Also contains the classifiers used in Brodu and Lague, 2012, as well as 2 vegetation classifiers for tidal environment and low-density long range scanners.

Zip - 1 Mo
Almost everything you need for classification, april 2013

Don’t forget to download the latest executables from Nicolas Brodu’s website, and to unpack them in the exe directory.

Sample data

To test the classification software using the Otira gorge case, you also need to download the following point clouds and place them in the Data directory.

Zip - 14.2 Mo
Otira Point Cloud Dataset

To test the tutorial on classifier construction using tidal vegetation environment, you need to download the data here :

Zip - 14.8 Mo
Tidal vegetation dataset

Dans la même rubrique :