Construct a convex hull polygon for a plane model

In this tutorial we will learn how to calculate a simple 2D convex hull polygon for a set of points supported by a plane.

The following video shows a demonstration of the code given below on the test dataset table_scene_mug_stereo_textured.pcd.

The code

First, download the dataset table_scene_mug_stereo_textured.pcd and save it somewhere to disk.

Then, create a file, let’s say, convex_hull_2d.cpp in your favorite editor, and place the following inside it:

 1#include <pcl/ModelCoefficients.h>
 2#include <pcl/io/pcd_io.h>
 3#include <pcl/point_types.h>
 4#include <pcl/sample_consensus/method_types.h>
 5#include <pcl/sample_consensus/model_types.h>
 6#include <pcl/filters/passthrough.h>
 7#include <pcl/filters/project_inliers.h>
 8#include <pcl/segmentation/sac_segmentation.h>
 9#include <pcl/surface/convex_hull.h>
10
11int
12 main ()
13{
14  pcl::PointCloud<pcl::PointXYZ>::Ptr cloud (new pcl::PointCloud<pcl::PointXYZ>), cloud_filtered (new pcl::PointCloud<pcl::PointXYZ>), cloud_projected (new pcl::PointCloud<pcl::PointXYZ>);
15  pcl::PCDReader reader;
16  reader.read ("table_scene_mug_stereo_textured.pcd", *cloud);
17
18  // Build a filter to remove spurious NaNs and scene background
19  pcl::PassThrough<pcl::PointXYZ> pass;
20  pass.setInputCloud (cloud);
21  pass.setFilterFieldName ("z");
22  pass.setFilterLimits (0, 1.1);
23  pass.filter (*cloud_filtered);
24  std::cerr << "PointCloud after filtering has: " << cloud_filtered->size () << " data points." << std::endl;
25
26  pcl::ModelCoefficients::Ptr coefficients (new pcl::ModelCoefficients);
27  pcl::PointIndices::Ptr inliers (new pcl::PointIndices);
28  // Create the segmentation object
29  pcl::SACSegmentation<pcl::PointXYZ> seg;
30  // Optional
31  seg.setOptimizeCoefficients (true);
32  // Mandatory
33  seg.setModelType (pcl::SACMODEL_PLANE);
34  seg.setMethodType (pcl::SAC_RANSAC);
35  seg.setDistanceThreshold (0.01);
36
37  seg.setInputCloud (cloud_filtered);
38  seg.segment (*inliers, *coefficients);
39
40  // Project the model inliers
41  pcl::ProjectInliers<pcl::PointXYZ> proj;
42  proj.setModelType (pcl::SACMODEL_PLANE);
43  proj.setInputCloud (cloud_filtered);
44  proj.setIndices (inliers);
45  proj.setModelCoefficients (coefficients);
46  proj.filter (*cloud_projected);
47
48  // Create a Convex Hull representation of the projected inliers
49  pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_hull (new pcl::PointCloud<pcl::PointXYZ>);
50  pcl::ConvexHull<pcl::PointXYZ> chull;
51  chull.setInputCloud (cloud_projected);
52  chull.reconstruct (*cloud_hull);
53
54  std::cerr << "Convex hull has: " << cloud_hull->size () << " data points." << std::endl;
55
56  pcl::PCDWriter writer;
57  writer.write ("table_scene_mug_stereo_textured_hull.pcd", *cloud_hull, false);
58
59  return (0);
60}

The explanation

The only interesting part is in the lines below, where the ConvexHull object gets created and the reconstruction is performed:

  pcl::PointCloud<pcl::PointXYZ>::Ptr cloud_hull (new pcl::PointCloud<pcl::PointXYZ>);
  pcl::ConvexHull<pcl::PointXYZ> chull;
  chull.setInputCloud (cloud_projected);
  chull.reconstruct (*cloud_hull);

Compiling and running the program

Add the following lines to your CMakeLists.txt file:

 1cmake_minimum_required(VERSION 3.5 FATAL_ERROR)
 2
 3project(convex_hull_2d)
 4
 5find_package(PCL 1.2 REQUIRED)
 6
 7include_directories(${PCL_INCLUDE_DIRS})
 8link_directories(${PCL_LIBRARY_DIRS})
 9add_definitions(${PCL_DEFINITIONS})
10
11add_executable (convex_hull_2d convex_hull_2d.cpp)
12target_link_libraries (convex_hull_2d ${PCL_LIBRARIES})

After you have made the executable, you can run it. Simply do:

$ ./convex_hull_2d

You will see something similar to:

PointCloud after filtering has: 139656 data points.
Convex hull has: 30 data points.