We supply two point cloud tiles for demonstration purposes, as well as the necessary data sources to run our code. The notebooks provide information on how to extract data for arbitrary point cloud tiles.
pointcloud/filtered_23**_970*.laz
Two 50x50m urban point clouds, courtesy of CycloMedia. See notes below for details.ahn/ahn_23**_970*.laz
The corresponding AHN point clouds, which can also be generated using the preprocessing tools.ahn/ahn_23**_970*.npz
The pre-processed ground and building surfaces from the AHN point clouds.bgt/bgt_buildings_demo.csv
BGT building footprint polygons for the two demo tiles.bgt/bgt_roads_demo.csv
BGT road part polygons of the parking spots and road surfaces in the area.bgt/custom_points_demo.csv
This contains <x,y> coordinates of pole-like objects and trees. These were collected manually from different sources, but can also be scraped from PDOK.
These files are sufficient to run the notebooks. Some additional required data files can be downloaded with provided scripts.
This repository was designed to be used with specific data sources:
- LAS point clouds of urban scenes supplied by CycloMedia.
- AHN3 point clouds downloaded from ArcGIS
- BGT data down from PDOK or the Amsterdam API.
The latter two sources are specific to The Netherlands.
We follow the naming conventions used by CycloMedia, which are based on tile codes. Each tile covers an area of exactly 50x50m, and is marked by the coordinates of the lower left corner following the Dutch Rijksdriehoeksstelsel + NAP (EPSG:7415).
The tile code is generated as follows:
tilecode = [X-coordinaat/50]_[Y-coordinaat/50]
In our example data files, this is 2386_9702 which thus translates to (119300, 485100) meters in RD coordinates or roughly (52.35264, 4.86321) degrees.