CUDARS: Angular Radon Spectrum - CUDA Version

Copyright (C) 2017-2020 Dario Lodi Rizzini.

Copyright (C) 2021- Dario Lodi Rizzini, Ernesto Fontana.

OVERVIEW

Library cudars implements the Angular Radon Spectrum method for estimation of rotation. It has been kept to a minimal design.

If you use this library, please cite the following paper:

Fontana E, Lodi Rizzini D. Accurate Global Point Cloud Registration Using GPU-Based Parallel Angular Radon Spectrum. Sensors. 2023; 23(20):8628. DOI 10.3390/s23208628.

@Article{FontanaLodiRizzini2023cudars,
	AUTHOR = {Fontana, Ernesto and Lodi Rizzini, Dario},
	TITLE = {Accurate Global Point Cloud Registration Using GPU-Based Parallel Angular Radon Spectrum},
	JOURNAL = {Sensors},
	VOLUME = {23},
	YEAR = {2023},
	NUMBER = {20},
	ARTICLE-NUMBER = {8628},
	URL = {https://www.mdpi.com/1424-8220/23/20/8628},
	ISSN = {1424-8220},
	DOI = {10.3390/s23208628}
}

or the most relevant associated publications by visiting: https://rimlab.ce.unipr.it/

DEPENDENCIES

The software depends on the following external libraries

• Boost (submodule lexical_cast)
• Point Cloud Library 1.8+
• NVCC CUDA Compiler; arch >= 52 (tested mainly on 75, 86). Please note that part of the algorithms have to run on a NVidia GPU.

Other dependencies are placed in directory thirdparty. Some examples require the external application "gnuplot" to display results. Necessary functions of common libraries (used also in the original ARS) such Eigen have been reimplemented in CUDA compatible code, generally usable in both host and device.

HOW TO COMPILE

Let ${cudars_ROOT} be the install directory of your local copy of library cudars. The following standard commands are required to compile it:

cd ${cudars_ROOT}
mkdir build
cd build
cmake ..
make

You can also install the library into a system directory. To change the install directory you must set cmake environment variable ${CMAKE_INSTALL_PREFIX} (e.g. using command "ccmake .." before calling "cmake .."). Its default value on UNIX-like/Linux systems is "/usr/local". After compiling library cudars, run the command:

sudo make install

The command "sudo" is required only if ${CMAKE_INSTALL_PREFIX} is a system diretory managed by administrator user root. Such command copies:

1. header files of

${cudars_ROOT}/include/cudars

  to

${CMAKE_INSTALL_PREFIX}/include/cudars/

2. library files

${cudars_ROOT}/lib/libcudars.a

  to

${CMAKE_INSTALL_PREFIX}/lib/

3. cmake script

${cudars_ROOT}/cmake_modules/cudarsConfig.cmake

  to

${CMAKE_INSTALL_PREFIX}/share/cudars/

HOW TO USE LIBRARY cudars IN YOUR PROJECT

If library cudars has been installed in system directory "/usr/local", then it is straighforward to use it in your projects. You need to add the following lines to your project as in this example:

CMAKE_MINIMUM_REQUIRED(VERSION 3.8)
PROJECT(foobar)

find_package(cudars REQUIRED)
message(STATUS "cudars_FOUND ${cudars_FOUND}")
message(STATUS "cudars_INCLUDE_DIRS ${cudars_INCLUDE_DIRS}")
message(STATUS "cudars_LIBRARY_DIRS ${cudars_LIBRARY_DIRS}")
message(STATUS "cudars_LIBRARIES ${cudars_LIBRARIES}")

if(${cudars_FOUND})
include_directories(${cudars_INCLUDE_DIRS})
link_directories(${cudars_LIBRARY_DIRS})
endif()

add_executable(foobar foobar.cpp) target_link_libraries(foobar ${cudars_LIBRARIES})

The above example uses the variables defined in cudarsConfig.cmake:

cudars_FOUND - system has cudars module
cudars_INCLUDE_DIRS - the cudars include directories
cudars_LIBRARY_DIRS - the cudars library directories
cudars_LIBRARIES - link these to use cudars