Installation

PCL Julia packages have fairly complex dependencies, so please take a careful look at the installation guide for each dependency.

Docker

If you don't want to build entire complex dependencies, you can use our docker image, PCL.jl and all dependencies installed. First install Docker, and then run:

docker pull r9y9/julia-cxx
docker run -it -rm r9y9/julia-cxx:PCL

Docker files are available at r9y9/julia-cxx.

If you want to build PCL.jl for yourself, read the following guides.

Requiements

Platform

The packages are tested on the following platforms:

macOS 10.11 (clang)
Ubuntu 14.04 (gcc-6/g++-6)

PCL

You need to install PCL 1.8 or later with shared library option ON. Since PCL have a fair amount of dependencies, be carefull of its dependencies, otherwise you might suffer from build issues.

macOS

On macOS, the following command will help to install PCL dependencies:

brew install cmake openni openni2 qhull boost glew flann eigen libusb vtk
export OPENNI2_INCLUDE=/usr/local/include/ni2
export OPENNI2_REDIST=/usr/local/lib/ni2

Ubuntu 14.04

For Ubuntu, there is a docker image that is confirmed to work with PCL as well as PCL.jl. Check out for the required apt packages from the docker files in r9y9/julia-cxx.

Compile and install PCL from source

If you have PCL dependencies installed, then recommended installation steps are as follows:

git clone https://github.com/PointCloudLibrary/pcl && cd pcl
mkdir build && cd build
cmake -DPCL_SHARED_LIBS=ON -DCMAKE_BUILD_TYPE=Release -DBUILD_global_tests=OFF -DBUILD_tools=OFF ..
make -j4
make -j4 install

Julia

Julia 0.5 or later is required. Install Julia v0.5 from binary distributions or build it from source.

You need to install r9y9/Cxx.jl [1]. Building Cxx.jl is a bit complex, as it builds its own llvm and clang by default. Please take a careful look at the installation guide of Cxx.jl. Ideally, the installation can be done as follows:

Pkg.clone("https://github.com/r9y9/Cxx.jl")
Pkg.build("Cxx")

[1]

Forked from Keno/Cxx.jl for RTTI support. Keno/Cxx.jl doesn't provides a way to enable RTTI without code modification for now.

Install PCL Julia packages

Set environmental variables property

There are a few environmental variables to be set property before installing Julia packages, otherwise it throws errors during package compilation time. You must tell locations of PCL dependencies (boost, FLANN, Eigen and VTK) to Julia via the following environmental variables:

Key	Default	Description
`BOOST_INCLUDE_DIR`	`/usr/local/include` (`/usr/include/` for linux)	Boost include directory
`FLANN_INCLUDE_DIR`	`/usr/local/include` (`/usr/include/` for linux)	Flann include directory
`EIGEN_INCLUDE_DIR`	`/usr/local/include/eigen3` (`/usr/include/eigen3` for linux)	Eigen include directory
`VTK_INCLUDE_DIR_PARENT`	`/usr/local/include` (`/usr/include/` for linux)	Parent directory for VTK includes
`VTK_INCLUDE_DIR`	`${VTK_INCLUDE_DIR_PARENT}/vtk-${major}.${minor}`	VTK include directory (`${major}` and `${minor}` will be automatically detected)

If you have installed all PCL dependencies into its default install path, you might not need to change the default values by setting environmental variables.

Clone and build

You are almost there! Add and clone the packages:

Pkg.add("BinDeps")
Pkg.add("DocStringExtensions")

Pkg.clone("https://github.com/JuliaPCL/PCLCommon.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLIO.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLFeatures.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLFilters.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLVisualization.jl")
Pkg.clone("https://github.com/JuliaPCL/LibPCL.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLKDTree.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLOctree.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLRecognition.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLSampleConsensus.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLSearch.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLSegmentation.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLTracking.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLRegistration.jl")
Pkg.clone("https://github.com/JuliaPCL/PCLKeyPoints.jl")
Pkg.clone("https://github.com/JuliaPCL/PCL.jl")

and then:

Pkg.build("LibPCL")

which searches installed PCL shared libraries. If it fails, please make sure again that you have set correct environmental variables. If you don't have PCL installed, Pkg.build("LibPCL") will try to build and install them into the LibPCL package directory but not recommended, unless if you have perfect requirements to build PCL.

Test if the installation succeeded

Pkg.test("PCL")

If it succeeded, installation is done. If you encounter errors even though all the previous steps succeeded, please file a bug report.