Downsampling 3D Point Clouds with a Voxelized Grid#

This tutorial shows how to use these optimizations inside a Docker* image. For the same functionality outside of Docker* images, see PCL Optimizations Outside of Docker* Images.

Prepare the environment:

cd <edge_insights_for_amr_path>/Edge_Insights_for_Autonomous_Mobile_Robots_<version>/AMR_containers
./run_interactive_docker.sh eiforamr-full-flavour-sdk:2023.1 root -c full_flavor
mkdir voxel_grid && cd voxel_grid

Create the file oneapi_voxel_grid.cpp:
```
vim oneapi_voxel_grid.cpp
```

Place the following inside the file:

#include <pcl/oneapi/filters/voxel_grid.h>
#include <pcl/io/pcd_io.h>
#include <pcl/point_types.h>
#include <pcl/point_cloud.h>


int main (int argc, char** argv)
{
  // Read Point Cloud
  pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_( new pcl::PointCloud<pcl::PointXYZ>() );
  int result = pcl::io::loadPCDFile("table_scene_lms400.pcd", *cloud_);
  if (result != 0)
  {
    pcl::console::print_info ("Load pcd file failed.\n");
    return result;
  }

  // Prepare Device Point Cloud Memory (input)
  pcl::oneapi::VoxelGrid::PointCloud_xyz cloud_device_xyz;
  cloud_device_xyz.upload(cloud_->points);
  pcl::oneapi::VoxelGrid::PointCloud & cloud_device = (pcl::oneapi::VoxelGrid::PointCloud &)cloud_device_xyz;

  // Prepare Device Point Cloud Memory (output)
  pcl::oneapi::VoxelGrid::PointCloud_xyz cloud_device_xyz_o;
  cloud_device_xyz_o.create(cloud_->size());
  pcl::oneapi::VoxelGrid::PointCloud & cloud_device_o = (pcl::oneapi::VoxelGrid::PointCloud &)cloud_device_xyz_o;

  // GPU calculate
  pcl::oneapi::VoxelGrid vg_oneapi;
  vg_oneapi.setInputCloud(cloud_device);
  float leafsize= 0.005f;
  vg_oneapi.setLeafSize (leafsize, leafsize, leafsize);
  vg_oneapi.filter(cloud_device_o);

  // print log
  std::cout << "[oneapi voxel grid] PointCloud before filtering: " << cloud_device.size() << std::endl;
  std::cout << "[oneapi voxel grid] PointCloud after filtering: " << cloud_device_o.size() << std::endl;
}

Create a CMakeLists.txt file:
```
vim CMakeLists.txt
```

Place the following inside the file:

cmake_minimum_required(VERSION 3.5 FATAL_ERROR)
set(target oneapi_voxel_grid)
set(CMAKE_CXX_COMPILER dpcpp)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_FLAGS "-Wall -Wpedantic -Wno-unknown-pragmas -Wno-pass-failed -Wno-unneeded-internal-declaration -Wno-unused-function -Wno-gnu-anonymous-struct -Wno-nested-anon-types -Wno-extra-semi -Wno-unused-local-typedef -fsycl -fsycl-unnamed-lambda -ferror-limit=1")
project(${target})

find_package(PCL 1.12 REQUIRED)
find_package(PCL-ONEAPI 1.12 REQUIRED)

include_directories(${PCL_INCLUDE_DIRS} ${PCL-ONEAPI_INCLUDE_DIRS})
link_directories(${PCL_LIBRARY_DIRS} ${PCL-ONEAPI_LIBRARY_DIRS})
add_definitions(${PCL_DEFINITIONS} ${PCL-ONEAPI_DEFINITIONS})

add_executable (${target} oneapi_voxel_grid.cpp)
target_link_libraries (${target} sycl pcl_oneapi_filters ${PCL_LIBRARIES})

Source the Intel® oneAPI Base Toolkit environment:

export PATH=/home/eiforamr/workspace/lib/pcl/share/pcl-1.12:/home/eiforamr/workspace/lib/pcl/share/pcl-oneapi-1.12:$PATH
source /opt/intel/oneapi/setvars.sh

Build the code:

cd /home/eiforamr/workspace/voxel_grid/
mkdir build && cd build
cmake ../
make -j

Download the test data from GitHub*:

wget https://raw.githubusercontent.com/PointCloudLibrary/data/5c26bdd0591ba150b91858b5c9fe5e91cb39ae86/tutorials/table_scene_lms400.pcd
# if the binary is not downloaded try setting the proxies first and try again:
export http_proxy="http://<http_proxy>:port"
export https_proxy="http://<https_proxy>:port"

Run the binary:
```
./oneapi_voxel_grid
```

Expected results example:

[oneapi voxel grid] PointCloud before filtering: 460400
[oneapi voxel grid] PointCloud after filtering: 41049

Code Explanation#

Load the test data from GitHub* into a PointCloud<PointXYZ>.

  // Read Point Cloud
  pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_( new pcl::PointCloud<pcl::PointXYZ>() );
  int result = pcl::io::loadPCDFile("table_scene_lms400.pcd", *cloud_);
  if (result != 0)
  {
    pcl::console::print_info ("Load pcd file failed.\n");
    return result;
  }

Create the GPU input and output device arrays, and upload the point cloud data to the input device array.

  // Prepare Device Point Cloud Memory (input)
  pcl::oneapi::VoxelGrid::PointCloud_xyz cloud_device_xyz;
  cloud_device_xyz.upload(cloud_->points);
  pcl::oneapi::VoxelGrid::PointCloud & cloud_device = (pcl::oneapi::VoxelGrid::PointCloud &)cloud_device_xyz;

  // Prepare Device Point Cloud Memory (output)
  pcl::oneapi::VoxelGrid::PointCloud_xyz cloud_device_xyz_o;
  cloud_device_xyz_o.create(cloud_->size());
  pcl::oneapi::VoxelGrid::PointCloud & cloud_device_o = (pcl::oneapi::VoxelGrid::PointCloud &)cloud_device_xyz_o;

The GPU starts to compute the model.

  // GPU calculate
  pcl::oneapi::VoxelGrid vg_oneapi;
  vg_oneapi.setInputCloud(cloud_device);
  float leafsize= 0.005f;
  vg_oneapi.setLeafSize (leafsize, leafsize, leafsize);
  vg_oneapi.filter(cloud_device_o);

Result (output log):

  // print log
  std::cout << "[oneapi voxel grid] PointCloud before filtering: " << cloud_device.size() << std::endl;
  std::cout << "[oneapi voxel grid] PointCloud after filtering: " << cloud_device_o.size() << std::endl;

Downsampling 3D Point Clouds with a Voxelized Grid

Contents

Downsampling 3D Point Clouds with a Voxelized Grid#

Code Explanation#