ITK/Examples/Statistics/KdTreeBasedKmeansEstimator3D

From KitwarePublic
Jump to navigationJump to search
ITK Examples Baseline Statistics TestKdTreeBasedKMeansClustering 3D.png

Description

Cluster a collection of measurements using the KMeans algorithm. The name "KdTreeBased" indicates that this is an efficient implementation which uses a KdTree.

ITK Classes Demonstrated

Output

The input is shown on the left. It consists of a single collection of 3D points that lend themselves to easy clustering into 2 clusters. The output clusters are shown on the right. Points belonging to the same cluster as shown in the same color.

KdTreeBasedKMeansClustering_3D.cxx

<source lang="cpp">

  1. include "itkDecisionRule.h"
  2. include "itkVector.h"
  3. include "itkListSample.h"
  4. include "itkKdTree.h"
  5. include "itkWeightedCentroidKdTreeGenerator.h"
  6. include "itkKdTreeBasedKmeansEstimator.h"
  7. if ITK_VERSION_MAJOR < 4
  8. include "itkMinimumDecisionRule2.h"
  9. else
  10. include "itkMinimumDecisionRule.h"
  11. endif
  12. include "itkEuclideanDistanceMetric.h"
  13. include "itkDistanceToCentroidMembershipFunction.h"
  14. include "itkSampleClassifierFilter.h"
  15. include "itkNormalVariateGenerator.h"
  1. include "vtkActor.h"
  2. include "vtkInteractorStyleTrackballCamera.h"
  3. include "vtkPolyData.h"
  4. include "vtkPolyDataMapper.h"
  5. include "vtkProperty.h"
  6. include "vtkRenderer.h"
  7. include "vtkRenderWindow.h"
  8. include "vtkRenderWindowInteractor.h"
  9. include "vtkSmartPointer.h"
  10. include "vtkVertexGlyphFilter.h"

int main(int, char *[]) { 

 typedef itk::Vector< double, 3 > MeasurementVectorType;
 typedef itk::Statistics::ListSample< MeasurementVectorType > SampleType;
 SampleType::Pointer sample = SampleType::New();

 typedef itk::Statistics::NormalVariateGenerator NormalGeneratorType;
 NormalGeneratorType::Pointer normalGenerator = NormalGeneratorType::New();

 normalGenerator->Initialize( 101 );

 MeasurementVectorType mv;
 double mean = 100;
 double standardDeviation = 30;
 for ( unsigned int i = 0 ; i < 100 ; ++i )
   {
   mv[0] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   mv[1] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   mv[2] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   sample->PushBack( mv );
   }

 normalGenerator->Initialize( 3024 );
 mean = 200;
 standardDeviation = 30;
 for ( unsigned int i = 0 ; i < 100 ; ++i )
   {
   mv[0] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   mv[1] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   mv[2] = ( normalGenerator->GetVariate() * standardDeviation ) + mean;
   sample->PushBack( mv );
   }

 typedef itk::Statistics::WeightedCentroidKdTreeGenerator< SampleType >
   TreeGeneratorType;
 TreeGeneratorType::Pointer treeGenerator = TreeGeneratorType::New();

 treeGenerator->SetSample( sample );
 treeGenerator->SetBucketSize( 16 );
 treeGenerator->Update();

 typedef TreeGeneratorType::KdTreeType TreeType;
 typedef itk::Statistics::KdTreeBasedKmeansEstimator<TreeType> EstimatorType;
 EstimatorType::Pointer estimator = EstimatorType::New();

 EstimatorType::ParametersType initialMeans(6);
 initialMeans[0] = 0.0; // Cluster 1, mean[0]
 initialMeans[1] = 0.0; // Cluster 1, mean[1]
 initialMeans[2] = 0.0; // Cluster 1, mean[2]
 initialMeans[3] = 5.0; // Cluster 2, mean[0]
 initialMeans[4] = 5.0; // Cluster 2, mean[1]
 initialMeans[5] = 5.0; // Cluster 2, mean[2]

 estimator->SetParameters( initialMeans );
 estimator->SetKdTree( treeGenerator->GetOutput() );
 estimator->SetMaximumIteration( 200 );
 estimator->SetCentroidPositionChangesThreshold(0.0);
 estimator->StartOptimization();

 EstimatorType::ParametersType estimatedMeans = estimator->GetParameters();

 for ( unsigned int i = 0 ; i < 6 ; i+=2 )
   {
   std::cout << "cluster[" << i << "] " << std::endl;
   std::cout << "    estimated mean : " << estimatedMeans[i] << " , " << estimatedMeans[i+1] << std::endl;
   }

 typedef itk::Statistics::DistanceToCentroidMembershipFunction< MeasurementVectorType >
   MembershipFunctionType;
 typedef MembershipFunctionType::Pointer                      MembershipFunctionPointer;
  1. if ITK_VERSION_MAJOR < 4
 typedef itk::Statistics::MinimumDecisionRule2 DecisionRuleType;
  1. else
 typedef itk::Statistics::MinimumDecisionRule DecisionRuleType;
  1. endif
 DecisionRuleType::Pointer decisionRule = DecisionRuleType::New();

 typedef itk::Statistics::SampleClassifierFilter< SampleType > ClassifierType;
 ClassifierType::Pointer classifier = ClassifierType::New();

 classifier->SetDecisionRule(decisionRule);
 classifier->SetInput( sample );
 classifier->SetNumberOfClasses( 2 );

 typedef ClassifierType::ClassLabelVectorObjectType               ClassLabelVectorObjectType;
 typedef ClassifierType::ClassLabelVectorType                     ClassLabelVectorType;
 typedef ClassifierType::MembershipFunctionVectorObjectType       MembershipFunctionVectorObjectType;
 typedef ClassifierType::MembershipFunctionVectorType             MembershipFunctionVectorType;

 ClassLabelVectorObjectType::Pointer  classLabelsObject = ClassLabelVectorObjectType::New();
 classifier->SetClassLabels( classLabelsObject );

 ClassLabelVectorType &  classLabelsVector = classLabelsObject->Get();
 classLabelsVector.push_back( 100 );
 classLabelsVector.push_back( 200 );



 MembershipFunctionVectorObjectType::Pointer membershipFunctionsObject =
   MembershipFunctionVectorObjectType::New();
 classifier->SetMembershipFunctions( membershipFunctionsObject );

 MembershipFunctionVectorType &  membershipFunctionsVector = membershipFunctionsObject->Get();

 MembershipFunctionType::CentroidType origin( sample->GetMeasurementVectorSize() );
 int index = 0;
 for ( unsigned int i = 0 ; i < 2 ; i++ )
   {
   MembershipFunctionPointer membershipFunction = MembershipFunctionType::New();
   for ( unsigned int j = 0 ; j < sample->GetMeasurementVectorSize(); j++ )
     {
     origin[j] = estimatedMeans[index++];
     }
   membershipFunction->SetCentroid( origin );
   membershipFunctionsVector.push_back( membershipFunction.GetPointer() );
   }

 classifier->Update();

 const ClassifierType::MembershipSampleType* membershipSample = classifier->GetOutput();
 ClassifierType::MembershipSampleType::ConstIterator iter = membershipSample->Begin();

 while ( iter != membershipSample->End() )
   {
   std::cout << "measurement vector = " << iter.GetMeasurementVector()
             << "class label = " << iter.GetClassLabel()
             << std::endl;
   ++iter;
   }

 // Visualize
 vtkSmartPointer<vtkPoints> points1 =
   vtkSmartPointer<vtkPoints>::New();
 vtkSmartPointer<vtkPoints> points2 =
   vtkSmartPointer<vtkPoints>::New();

 iter = membershipSample->Begin();
 while ( iter != membershipSample->End() )
   {
   if(iter.GetClassLabel() == 100)
     {
     points1->InsertNextPoint(
       iter.GetMeasurementVector()[0],
       iter.GetMeasurementVector()[1],
       iter.GetMeasurementVector()[2]);
     }
   else
     {
     points2->InsertNextPoint(
       iter.GetMeasurementVector()[0],
       iter.GetMeasurementVector()[1],
       iter.GetMeasurementVector()[2]);
     }
   ++iter;
   }

 vtkSmartPointer<vtkPolyData> polyData1 =
   vtkSmartPointer<vtkPolyData>::New();
 polyData1->SetPoints(points1);
 vtkSmartPointer<vtkVertexGlyphFilter> glyphFilter1 =
   vtkSmartPointer<vtkVertexGlyphFilter>::New();
 glyphFilter1->SetInputConnection(polyData1->GetProducerPort());
 glyphFilter1->Update();
 vtkSmartPointer<vtkPolyDataMapper> mapper1 =
   vtkSmartPointer<vtkPolyDataMapper>::New();
 mapper1->SetInputConnection(glyphFilter1->GetOutputPort());
 vtkSmartPointer<vtkActor> actor1 =
   vtkSmartPointer<vtkActor>::New();
 actor1->GetProperty()->SetColor(0,1,0);
 actor1->GetProperty()->SetPointSize(3);
 actor1->SetMapper(mapper1);

 vtkSmartPointer<vtkPolyData> polyData2 =
   vtkSmartPointer<vtkPolyData>::New();
 polyData2->SetPoints(points2);
 vtkSmartPointer<vtkVertexGlyphFilter> glyphFilter2 =
   vtkSmartPointer<vtkVertexGlyphFilter>::New();
 glyphFilter2->SetInputConnection(polyData2->GetProducerPort());
 glyphFilter2->Update();
 vtkSmartPointer<vtkPolyDataMapper> mapper2 =
   vtkSmartPointer<vtkPolyDataMapper>::New();
 mapper2->SetInputConnection(glyphFilter2->GetOutputPort());
 vtkSmartPointer<vtkActor> actor2 =
   vtkSmartPointer<vtkActor>::New();
 actor2->GetProperty()->SetColor(1,0,0);
 actor2->GetProperty()->SetPointSize(3);
 actor2->SetMapper(mapper2);

 vtkSmartPointer<vtkRenderWindow> renderWindow =
   vtkSmartPointer<vtkRenderWindow>::New();
 renderWindow->SetSize(300,300);

 vtkSmartPointer<vtkRenderer> renderer =
   vtkSmartPointer<vtkRenderer>::New();
 renderWindow->AddRenderer(renderer);

 renderer->AddActor(actor1);
 renderer->AddActor(actor2);
 renderer->ResetCamera();
 renderer->Render();

 vtkSmartPointer<vtkRenderWindowInteractor> renderWindowInteractor =
   vtkSmartPointer<vtkRenderWindowInteractor>::New();
 vtkSmartPointer<vtkInteractorStyleTrackballCamera> style =
   vtkSmartPointer<vtkInteractorStyleTrackballCamera>::New();

 renderWindowInteractor->SetInteractorStyle(style);

 renderWindowInteractor->SetRenderWindow(renderWindow);
 renderWindowInteractor->Initialize();

 renderWindowInteractor->Start();

 return EXIT_SUCCESS;

} </source>


CMakeLists.txt

<syntaxhighlight lang="cmake"> cmake_minimum_required(VERSION 3.9.5)

project(KdTreeBasedKMeansClustering_3D)

find_package(ITK REQUIRED) include(${ITK_USE_FILE}) if (ITKVtkGlue_LOADED) 

 find_package(VTK REQUIRED)
 include(${VTK_USE_FILE})

else() 

 find_package(ItkVtkGlue REQUIRED)
 include(${ItkVtkGlue_USE_FILE})
 set(Glue ItkVtkGlue)

endif()

add_executable(KdTreeBasedKMeansClustering_3D MACOSX_BUNDLE KdTreeBasedKMeansClustering_3D.cxx) target_link_libraries(KdTreeBasedKMeansClustering_3D 

 ${Glue}  ${VTK_LIBRARIES} ${ITK_LIBRARIES})

</syntaxhighlight>

Download and Build KdTreeBasedKMeansClustering_3D

Click here to download KdTreeBasedKMeansClustering_3D. and its CMakeLists.txt file. Once the tarball KdTreeBasedKMeansClustering_3D.tar has been downloaded and extracted, 

cd KdTreeBasedKMeansClustering_3D/build
  • If ITK is installed:
cmake ..
  • If ITK is not installed but compiled on your system, you will need to specify the path to your ITK build:
cmake -DITK_DIR:PATH=/home/me/itk_build ..

Build the project, 

make

and run it: 

./KdTreeBasedKMeansClustering_3D

WINDOWS USERS PLEASE NOTE: Be sure to add the VTK and ITK bin directories to your path. This will resolve the VTK and ITK dll's at run time.

Building All of the Examples

Many of the examples in the ITK Wiki Examples Collection require VTK. You can build all of the the examples by following these instructions. If you are a new VTK user, you may want to try the Superbuild which will build a proper ITK and VTK.

ItkVtkGlue

ITK >= 4

For examples that use QuickView (which depends on VTK), you must have built ITK with Module_ITKVtkGlue=ON.

ITK < 4

Some of the ITK Examples require VTK to display the images. If you download the entire ITK Wiki Examples Collection, the ItkVtkGlue directory will be included and configured. If you wish to just build a few examples, then you will need to download ItkVtkGlue and build it. When you run cmake it will ask you to specify the location of the ItkVtkGlue binary directory.

Retrieved from "https://public.kitware.com/Wiki/index.php?title=ITK/Examples/Statistics/KdTreeBasedKmeansEstimator3D&oldid=43939"