Pointcloud Raster

Provides a simple tool wit no dependencies to generate a raster from a pointcloud.

Background

It is possible to create raster from PNGs using things like awesome cloudcompare. However, tools that do raster usually have many requirements to build and possibly license constraints. I wanted something simple for my QA needs, where I simply pass the cloud, takes the photos for me, and works in a headless environment.

In addition, I wanted to be able to do it on any device (IOS, android).

So, if you depend on high accuracy renders for your customers, this is not for you. But if you only need to create screenshots/rasters which are for previews or quick screenshots, enjoy.

Features

• Full support TXT and LAS files
• Rudimentary PLY support: Only XYZ and XYZRGB pointclouds
• Can output to PNG file (requires libpng)
• Supports rasters from arbitrary view points (cameras)
• Files are rasterized on fly, meaning low memory usage. However, raster is hold in memory, which implies max size depends on your system.

How to build

This program has no dependencies, beyond standard library, a modern CMake and compiler. Do:

git clone https://github.com/manlito/pointcloud-raster
mkdir -p pointcloud-raster-build
cd pointcloud-raster-build
cmake  ../pointcloud-raster
make -j4
make install

To disable example app and/or test, use these options. For instance:

cmake -D POINTCLOUD_RASTER_BUILD_APP=Off  -D POINTCLOUD_RASTER_BUILD_TESTS=Off ../pointcloud-raster

Without PNG

Such use case can be cool if you only need to return an in memory raster and have another way to display.

cmake -D POINTCLOUD_RASTER_BUILD_APP=Off -D POINTCLOUD_RASTER_PNG_SUPPORT=Off ../pointcloud-raster

Without LAS

If your use case needs a different way of reading, you should create your own PointcloudProvider. Please see TXTReader to see how. When, you should build as follows:

cmake -D POINTCLOUD_RASTER_BUILD_APP=Off -D POINTCLOUD_RASTER_LAS_SUPPORT=Off ../pointcloud-raster

How to use

Example app

We ship an app with the following supported parameters:

pointcloud_raster <cloud.las|cloud.txt> <LAS|TXT> <SIDE|FRONT|TOP|BOTTOM|PERSPECTIVE|ALL> <output_prefix> <max_size>

For instance:

pointcloud_raster cloud.las LAS TOP preview 1024

Creates a single file in working dir named preview_top.png. But special case:

pointcloud_raster cloud.las LAS ALL preview 1024

Creates 4 files in working dir.

Note: Output dir is expected to exist.

Library usage

First, you install it and then in your CMakeLists.txt search as:

find_package(PointcloudRaster 1.0 REQUIRED)
add_executable(your_program your_code.cpp)
target_link_libraries(your_program PointcloudRaster::PointcloudRaster)

See Examples repo for more.

Using in production

I can tell I have run this tool on pointcloud files with more than 500 million points without any problems. Indeed, I have tested more than 50 different pointclouds, and works as expected. No benchmarks, but certainly it is pretty fast. You should try it. Building is easy, specially with Dockerfile