Potree Point Clouds in the Browser with WebGL Daniel Kastl - FOSS4G - - PowerPoint PPT Presentation

Potree

Point Clouds in the Browser with WebGL

About me

Geographer, Mapper, Software Developer Maintainer of the pgRouting Project Founder of Georepublic Living in Germany and Japan Enjoy Open Source FOSS4G Recently involved in Point Clouds

What do we do with Point Clouds?

You can create your own Point Cloud?

1.2 GB LAS file ~45 Mio. points

How do we show mBillions of Points?

• A WebGL point cloud viewer for large datasets.
• An Open Source project developed by Markus Schütz (and

growing number of contributors)

• Financially supported by Rapidlasso, Georepublic and
• thers

Potree

• Only a small part of

data is essential for a given point of view

• A multi-resolution

Octree allows to quickly cull away unnecessary data and only load what is needed.

How does Potree work?

• Low-level nodes contain

low resolution models of large areas

• Resolution increases

while coverage decreases at higher levels

• Works similar to map

tiling

How does Potree work?

• Each Octree node stored

in a separate file

• Server is only necessary

as file store

• No server side

application required

How does Potree work?

Potree Javascript Library

• Data conversion with the PotreeConverter
• Uses Three.js rendering engine and inherits its

capabilities

• Client-side LAS and LAZ support from Plas.io
• API for (measurement) tools
• Open Source license (FreeBSD)

Data format (PotreeConverter)

• Input:

○ las, laz and ply files

• Output:

○ Cloud.js file for metadata (BBOX, spacing, storage format) ○ Octree nodes (las, laz or bin format) with position, color, intensity and classification information ○ hrc files for the Octree hierarchy

Storage Format

• Octree nodes stored in a directory hierarchy
• Each directory contains:

○ a subset of the Octree with 4-6 levels ○ a “hrc” file with the hierarchy, that is stored in the current directory

“Beautiful Point Clouds”

• Interpolation and high-

quality splats

○ Increase readability of fine details ○ Avoid overlapping points through nearest-neighbor like interpolation or by blending overlapping points together

“Beautiful Point Clouds”

• Eye-Dome-Lighting (EDL)

○ Illumination without normals ○ Outlines enhance depth- perception ○ Especially useful to render elevation, classification

• r single-color

Features at a Glance

• Support for very large point clouds
• Adaptive point size - adjust point size to reduce holes
• Eye-Dome-Lighting (EDL) - illumination without normals
• Renders RGB, elevation, intensity, classification, etc.
• High-quality rendering modes (interpolation & splats)
• Distance and area measurements, height profiles
• Clipping volumes

Limitations

• Browser security policies require to run Potree from a

web-server.

• Requires WebGL.

Some systems (Nvidia with Optimus) choose slow onboard graphics card for WebGL apps, even if a faster one is available. Fast one has to be chosen manually in NVIDIA settings.

• Only for viewing and analyzing, not manipulation of data.

Challenges

• Processing of large datasets
• Data streaming to the browser

○ 3D-Tiles specification?

• User-friendly navigation

Who is using Potree?

• Various Companies and Institutions
• As well as researchers, students, digital heritage, ...

Examples

http://potree.org/wp/demo

Version 1.3 released - What comes next?

• Projecting maps onto point clouds in real time
• Combine with the Cesium Globe renderer
• Load and display point cloud formats without conversion
• Further improve performance and push the boundaries
• Improve support for mobile devices
• Additional tools and rendering modes

A lot more is possible and funding will help to make it happen!

Interested in Potree?

Project: http://www.potree.org Email: daniel@georepublic.de Twitter: @dkastl

Thank you!