Cesium Terrain Builder Docker

This repo contains a Dockerfile for the Cesum Terrain Builder (CTB) app with support for the new Cesium terrain format quantized-mesh. It is build from a fork providing quantized-mesh support, as described in this artice. Information on the most recent development of this fork is available in this pull request. Thanks to @homme and @ahuarte47 for the great work on Cesium Terrain Builder and quantized-mesh support.

Note: The images are manually rebuild, when new commits are published at ahuarte47/cesium-terrain-builder/tree/master-quantized-mesh. If you miss an update in an image, please let us know by creating an issue.

If you experience problems or want to contribute please create an issue or pull request.

Follow the steps below to create your own quantized-mesh tiles for Cesium using this Docker image.

News

• 2023-08-06:

  • The additional arm64 image ins now available from both Dockerhub and Github packages.

    docker pull ghcr.io/tum-gis/ctb-quantized-mesh:latest

    docker pull ghcr.io/tum-gis/ctb-quantized-mesh:alpine

• 2023-07-31:

  • An additional arm64 version of the image was added. This currently untested and I'm happy for any feedback on this, see tum-gis/cesium-terrain-builder-docker#19. The arm64 Docker images are hosted in Github packages.

    docker pull ghcr.io/tum-gis/ctb-quantized-mesh:latest

    docker pull ghcr.io/tum-gis/ctb-quantized-mesh:alpine

• 2023-03-06:

  • Updated both images to reflect latest changes in ahuarte47/cesium-terrain-builder:master-quantized-mesh
  • Updated GDAL to 2.4.0 in latest image

• 2020-11: Updated alpine image to Alpine v3.12 and GDAL v3.14

• 2020-11: Reduced size of all images using multi stage builds.

Image variants

The amd64 Docker images are available on DockerHub from tumgis or from Github packages. To get the image run:

docker pull tumgis/ctb-quantized-mesh:<TAG> or docker pull ghcr.io/tum-gis/ctb-quantized-mesh:<TAG>

The arm64 Docker images are ONLY available from Github packages. To get the image run:

docker pull ghcr.io/tum-gis/ctb-quantized-mesh:<TAG>

Following tags are available:

Tag	Build status	Arch	Description
latest		amd64 arm64	Latest image build based on Debian and GDAL 2.4.0
alpine		amd64 arm64	Image based on leightweight Alpine Linux v3.12 and GDAL v3.14

Content

• Cesium Terrain Builder Docker
  • News
  • Image variants
  • Content
  • Preparation
    • Docker settings
    • Data pre-processing
    • Data storage
  • Cesium Terrain Builder usage
    • Start CTB container and mount data folder
      • Linux - bash
      • Windows - cmd
      • Windows - git-bash
      • Windows - powershell
    • Create a GDAL Virtual Dataset (optional)
    • Create Cesium Terrain files
    • Create Cesium layer description file
  • Troubleshooting
    • Performance issues
    • Handling large datasets

Preparation

Docker settings

The system resources Docker can use are limited by default on Windows systems. Goto Docker tray Icon -> Settings -> Advanced to adjust the number of cores and main memory Docker can use to increase performance.

Data pre-processing

It is highly recommended (but not required) to transform your data to the WGS84 (EPSG:4326) coordinate reference system before using CTB. This helps to avoid vertial or horizontal offsets of terrain datasets. Use the NTv2 transformation method if available. This is e.g. supported by FME using the EsriReprojector transformer or ESRI ArcGIS.

Data storage

Put your data in a folder, that can be mounted by Docker. On Windows, you will have to grant access to the drive where the data is located before being able to mount the folder. Goto Docker tray Icon -> Settings -> Shared Drives to share drives with Docker. Visit this blog post for a comprehensive guide on mounting host directories on Windows.

In the following we assume that your terrain data is stored in d:\docker\terrain for a Windows Docker host and drive d:\ is shared with Docker. For a Linux Docker host we assume your data is stored in /docker/terrain.

Cesium Terrain Builder usage

When your data is transformed and copied to a location available for Docker your are ready for creating a Cesium terrain with CTB.

Start CTB container and mount data folder

Before starting CTB it is recommended to pull the latest image version using docker pull tumgis/ctb-quantized-mesh. After that, start a CTB container and mount your terrain data folder to /data in the container. Follow the examples below for different operating systems and shells.

Linux - bash

docker run -it --name ctb \
    -v "/docker/terrain:/data" \
  tumgis/ctb-quantized-mesh

Windows - cmd

docker run -it --name ctb ^
    -v "d:/docker/terrain:/data" ^
  tumgis/ctb-quantized-mesh

Windows - git-bash

winpty docker run --rm -it --name ctb \
    -v "d:\\docker\\terrain:/data" \
  tumgis/ctb-quantized-mesh

Windows - powershell

docker run -it --name ctb `
    -v "d:\docker\terrain:/data" `
  tumgis/ctb-quantized-mesh

Create a GDAL Virtual Dataset (optional)

If you dataset consists of a single file, continue to the next step. If your dataset consists of multiple tiles (more than one file), a GDAL Virtual Dataset needs to be created using the gdalbuildvrt app.

gdalbuildvrt <output-vrt-file.vrt> <files>

For instance, if you have several *.tif files, run:

gdalbuildvrt tiles.vrt *.tif

More options to create a GDAL Virtual Dataset e.g. using a list of files are described in the gdalbuildvrt documentation.

Create Cesium Terrain files

First, create an output folder for you terrain, e.g. mkdir -p terrain. Second, run CTB to create the terrain files:

ctb-tile -f Mesh -C -N -o terrain <inputfile.tif or input.vrt>

For example, if a tile.vrt has been created as described above:

ctb-tile -f Mesh -C -N -o terrain tile.vrt

The ctb-tile app supports several options. Run ctb-tile --help to display all options. For larger datasets consider setting the -m option and the GDAL_CHACHEMAX environment variable as described here.

Create Cesium layer description file

Finally, a layer description file needs to be created. Simply run the same command you used for creating the terrain files again adding the -l switch. For instance:

ctb-tile -f Mesh -C -N -o terrain tiles.vrt            # Create terrain files
ctb-tile -f Mesh -C -N -l -o terrain tiles.vrt         # Create layer description file

Finally, your terrain data folder should look similar to this:

$ tree -v -C -L 1 terrain/
terrain/
|-- 0
|-- 1
|-- 2
|-- 3
|-- 4
|-- 5
|-- 6
|-- 7
|-- 8
|-- 9
|-- 10
|-- 11
|-- 12
|-- 13
|-- 14
|-- 15
`-- layer.json

The quantized-mesh terrain is now ready for usage.

Troubleshooting

Performance issues

Read the recommendations for ctb-tile carefully, especially when handling large datasets.

Handling large datasets

Datasets with a big extent can lead to overflow errors on lower zoom levels:

0...10...20...30...40...50...60...70...80...90...ERROR 1: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585
ERROR 1: IReadBlock failed at X offset 0, Y offset 0: IReadBlock failed at X offset 0, Y offset 0: Integer overflow : nSrcXSize=41494, nSrcYSize=16585

As described here, this is caused by GDAL trying to create overviews from input data. A possible solution is to create simplified versions of the input data with lower resolutions and use them for creating the mesh tiles on lower levels. This can be done using e.g. gdal_translate. After that, try to create mesh tiles using ctb-tile with the resolutions that do not crash starting from level 0. Try to use the highest resolution possible that does not crash for each level.