Application: gvSIG desktop » gvSIG tools: ToolBox

package es.unex.sextante.vectorTools.vectorSpatialCluster;

import java.util.Arrays;

import com.vividsolutions.jts.geom.Coordinate;

import es.unex.sextante.additionalInfo.AdditionalInfoNumericalValue;

import es.unex.sextante.additionalInfo.AdditionalInfoVectorLayer;

import es.unex.sextante.core.GeoAlgorithm;

import es.unex.sextante.core.Sextante;

import es.unex.sextante.dataObjects.IFeature;

import es.unex.sextante.dataObjects.IFeatureIterator;

import es.unex.sextante.dataObjects.IVectorLayer;

import es.unex.sextante.dataObjects.vectorFilters.BoundingBoxFilter;

import es.unex.sextante.exceptions.GeoAlgorithmExecutionException;

import es.unex.sextante.exceptions.RepeatedParameterNameException;

import es.unex.sextante.outputs.IOutputChannel;

import es.unex.sextante.outputs.Output;

import es.unex.sextante.outputs.OutputVectorLayer;

import es.unex.sextante.shapesTools.ShapesTools;

public class VectorSpatialClusterAlgorithm

         extends

            GeoAlgorithm {

   public static final String RESULT        = "RESULT";

   public static final String NUMCLASS      = "NUMCLASS";

   public static final String LAYER         = "LAYER";

   private ValueAndClass[]    m_Classes;

   private double             m_dMean[][];

   private int                m_iClasses;

   private int                m_iThreshold;

   private IVectorLayer       m_LayerIn;

   private static double      NO_DATA       = Double.NEGATIVE_INFINITY;

   private static int         NO_DATA_CLASS = Integer.MAX_VALUE;

   @Override

   public boolean processAlgorithm() throws GeoAlgorithmExecutionException {

      int i;

      final Class[] types = { Integer.class };

      final String[] sFields = { Sextante.getText("Class") };

      m_iClasses = m_Parameters.getParameterValueAsInt(NUMCLASS);

      m_LayerIn = m_Parameters.getParameterValueAsVectorLayer(LAYER);

      if (!m_bIsAutoExtent) {

         m_LayerIn.addFilter(new BoundingBoxFilter(m_AnalysisExtent));

      }

      m_Classes = new ValueAndClass[m_LayerIn.getShapesCount()];

      final IFeatureIterator iter = m_LayerIn.iterator();

      i = 0;

      while (iter.hasNext()) {

         final IFeature feature = iter.next();

         final Coordinate coord = feature.getGeometry().getCoordinate();

         m_Classes[i] = new ValueAndClass(2);

         m_Classes[i].dValue[0] = coord.x;

         m_Classes[i].dValue[1] = coord.y;

         i++;

      }

      classify();

      final Object[][] values = new Object[1][m_LayerIn.getShapesCount()];

      for (i = 0; i < m_Classes.length; i++) {

         values[0][i] = new Integer(m_Classes[i].iClass);

      }

      final IOutputChannel channel = getOutputChannel(RESULT);

      final Output out = new OutputVectorLayer();

      out.setDescription(Sextante.getText("Result"));

      out.setName(RESULT);

      out.setOutputChannel(channel);

      out.setOutputObject(ShapesTools.addFields(m_OutputFactory, m_LayerIn, channel, sFields, values, types));

      addOutputObject(out);

      return !m_Task.isCanceled();

   }

   @Override

   public void defineCharacteristics() {

      setName(Sextante.getText("Spatial_cluster"));

      setGroup(Sextante.getText("Tools_for_point_layers"));

      setUserCanDefineAnalysisExtent(true);

      try {

         m_Parameters.addInputVectorLayer(LAYER, Sextante.getText("Layer"), AdditionalInfoVectorLayer.SHAPE_TYPE_POINT, true);

         m_Parameters.addNumericalValue(NUMCLASS, Sextante.getText("Number_of_classes"),

                  AdditionalInfoNumericalValue.NUMERICAL_VALUE_INTEGER, 3, 2, Integer.MAX_VALUE);

         addOutputVectorLayer(RESULT, Sextante.getText("Result"), OutputVectorLayer.SHAPE_TYPE_POINT);

      }

      catch (final RepeatedParameterNameException e) {

         Sextante.addErrorToLog(e);

      }

   }

   private void initValues() {

      int i;

      int iValues = 0;

      boolean bNoData;

      double dStep;

      double dValue;

      final double dMin[] = new double[2];

      final double dMax[] = new double[2];

      for (i = 0; i < 2; i++) {

         dMin[i] = Double.MAX_VALUE;

         dMax[i] = Double.NEGATIVE_INFINITY;

      }

      for (i = 0; i < m_Classes.length; i++) {

         bNoData = false;

         for (int j = 0; j < m_Classes[i].dValue.length; j++) {

            dValue = m_Classes[i].dValue[j];

            if (dValue != NO_DATA) {

               dMin[j] = Math.min(dMin[j], dValue);

               dMax[j] = Math.max(dMax[j], dValue);

            }

            else {

               bNoData = true;

            }

         }

         if (bNoData) {

            m_Classes[i].iClass = NO_DATA_CLASS;

         }

         else {

            iValues++;

            m_Classes[i].iClass = 0;

         }

      }

      m_dMean = new double[m_iClasses][2];

      for (i = 0; i < 2; i++) {

         dStep = (dMax[i] - dMin[i]) / ((m_iClasses + 1));

         for (int j = 0; j < m_iClasses; j++) {

            m_dMean[j][i] = dMin[i] + dStep * (j + 1);

         }

      }

      m_iThreshold = (int) (iValues * 0.02);

   }

   private boolean classify() {

      int i, j;

      int iChangedCells;

      int iPrevClass;

      int iClass;

      final int iCells[] = new int[m_iClasses];

      double dNewMean[][];

      double swap[][];

      initValues();

      dNewMean = new double[m_iClasses][2];

      do {

         Arrays.fill(iCells, 0);

         iChangedCells = 0;

         for (i = 0; i < m_iClasses; i++) {

            Arrays.fill(dNewMean[i], 0.0);

         }

         for (i = 0; i < m_Classes.length; i++) {

            iPrevClass = m_Classes[i].iClass;

            if (iPrevClass != NO_DATA_CLASS) {

               iClass = getClass(m_Classes[i].dValue);

               m_Classes[i].iClass = iClass;

               for (j = 0; j < 2; j++) {

                  dNewMean[iClass][j] += m_Classes[i].dValue[j];

               }

               iCells[iClass]++;

               if (iClass != iPrevClass) {

                  iChangedCells++;

               }

            }

         }

         for (i = 0; i < 2; i++) {

            for (j = 0; j < m_iClasses; j++) {

               dNewMean[j][i] /= iCells[j];

            }

         }

         swap = m_dMean;

         m_dMean = dNewMean;

         dNewMean = swap;

         setProgressText(Sextante.getText("Modified_classes") + Integer.toString(iChangedCells));

         if (m_Task.isCanceled()) {

            return false;

         }

      }

      while (iChangedCells > m_iThreshold);

      return true;

   }

   private int getClass(final double[] dValues) {

      int iClass = 0;

      double dMinDist = Double.MAX_VALUE;

      double dDist;

      double dDif;

      for (int i = 0; i < m_iClasses; i++) {

         dDist = 0;

         for (int j = 0; j < dValues.length; j++) {

            dDif = m_dMean[i][j] - dValues[j];

            dDist += (dDif * dDif);

         }

         if (dDist < dMinDist) {

            dMinDist = dDist;

            iClass = i;

         }

      }

      return iClass;

   }

   private class ValueAndClass {

      public double dValue[];

      public int    iClass;

      public ValueAndClass(final int i) {

         dValue = new double[i];

      }

   }

}