@geodanresearch/mapbox-3dtiles

mapbox-3dtiles

3D Tiles implementation using Mapbox GL JS custom layers

See https://geodan.github.io/mapbox-3dtiles/ for a working demo.

This is a proof-of-concept implementation of a 3D Tiles viewer as a Mapbox GL JS custom layer. WebGL rendering is implemented using three.js. Only Web Mercator (EPSG:3857) tilesets are supported, as this is the projection mapbox uses. Earth-centered earth-fixed tilesets are explicitly not supported. Tilesets used for testing were generated using pg2b3dm, using a PostGIS database with EPSG:3857 geometries.

This is by no means a complete implementation of the 3D Tile specification. Currently the following features are supported:

• Geometric error based tile loading
• Replacement and additive refinement
• Only Box bounding volumes are supported
• Tile transforms
• External tilesets
• Tile types:
  • Batched 3D Model (b3dm)
  • Point Cloud (pnts): basic implementation
• Draco compressed i3dm and b3dm tiles

The following features are not supported at this time:

• Any coordinate system other than EPSG:3857
• Region and sphere bounding volumes
• Viewer request volumes
• Instanced 3D Model (i3dm) tiles
• Composite (cmpt) tiles
• 3D Tile Styles

Instructions

In a directory on your webserver run the folowing commands:

git clone https://github.com/Geodan/mapbox-3dtiles.git
cd mapbox-3dtiles
npm install
tar xvf data.tar.bz2

Next, copy file "apikeys.js.example" to "apikeys.js" and add your mapbox token.

You can then start the development environment:

npm start

Point your browser to http://localhost:8082 for an overview page of demo's.

Draco compression

Draco compressed data can be used with Mapbox-3DTiles, to do this create an pass an Threejs dracoloader to the layer.

import { DRACOLoader } from '../node_modules/three/examples/jsm/loaders/DRACOLoader.js';

var dracoLoader = new DRACOLoader();
dracoLoader.setDecoderPath("https://www.gstatic.com/draco/versioned/decoders/1.4.1/");

const myLayer = new Mapbox3DTiles.Mapbox3DTilesLayer(
	{
		id: 'my_awesome_layer',
		url: 'https://my-layer/tileset.json',
		dracoLoader: dracoLoader
	}
);

Creating tilesets

Tilesets can be created using pg2b3dm, using a PostGIS database table as source. The PostGIS table should contain 3D geometries in EPSG:3857 projection.

Example query creating extruded 3D buildings in EPSG:3857:

DROP TABLE IF EXISTS <schema>.<output_table>;
CREATE TABLE <schema>.<output_table> AS (
	WITH extent AS (
		SELECT ST_MakeEnvelope(<minx>, <miny>, <maxx>, <maxy>, <input_srid>) geom
	),
	footprints AS (
		SELECT a.id AS id, a.height, a.geom
		FROM <schema>.<input_table> a, extent b
		WHERE ST_Intersects(a.geom, b.geom)
	)
	SELECT id, ST_Force3D(ST_Extrude(ST_Transform(ST_MakeValid(geom), 3857), 0, 0, height)) AS geom
	FROM footprints
);
DELETE FROM <schema>.<output_table> WHERE geom IS NULL; -- cleanup
DELETE FROM <schema>.<output_table> WHERE ST_GeometryType(geom) NOT LIKE 'ST_PolyhedralSurface'; -- cleanup

Creating tileset using pg2b3dm:

pg2b3dm -h <my_host> -U <my_user> -d <my_database> -p <my_port> -c <geom_column> -t <my_schema.my_table>

For more information, see the pg2b3dm documentation.

Creating tileset from point cloud:

Pointcloud data is not yet supported by pg2b3dm and needs to be exporterd with an earlier tool called py3dtiles

py3dtiles convert --srs_in <srs_in> --srs_out 3857 --out <tileset_name> pointcloud.las

For more information, see the py3dtiles documentation.