root / branches / v2_0_0_prep / extensions / extRemoteSensing / src / org / gvsig / remotesensing / classification / ClassificationMaximumLikelihoodProcess.java @ 31496
History | View | Annotate | Download (12.5 KB)
| 1 |
/* gvSIG. Sistema de Informaci?n Geogr?fica de la Generalitat Valenciana
|
|---|---|
| 2 |
*
|
| 3 |
* Copyright (C) 2006 Instituto de Desarrollo Regional and Generalitat Valenciana.
|
| 4 |
*
|
| 5 |
* This program is free software; you can redistribute it and/or
|
| 6 |
* modify it under the terms of the GNU General Public License
|
| 7 |
* as published by the Free Software Foundation; either version 2
|
| 8 |
* of the License, or (at your option) any later version.
|
| 9 |
*
|
| 10 |
* This program is distributed in the hope that it will be useful,
|
| 11 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 12 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
| 13 |
* GNU General Public License for more details.
|
| 14 |
*
|
| 15 |
* You should have received a copy of the GNU General Public License
|
| 16 |
* along with this program; if not, write to the Free Software
|
| 17 |
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,USA.
|
| 18 |
*
|
| 19 |
* For more information, contact:
|
| 20 |
*
|
| 21 |
* Generalitat Valenciana
|
| 22 |
* Conselleria d'Infraestructures i Transport
|
| 23 |
* Av. Blasco Iba?ez, 50
|
| 24 |
* 46010 VALENCIA
|
| 25 |
* SPAIN
|
| 26 |
*
|
| 27 |
* +34 963862235
|
| 28 |
* gvsig@gva.es
|
| 29 |
* www.gvsig.gva.es
|
| 30 |
*
|
| 31 |
* or
|
| 32 |
*
|
| 33 |
* Instituto de Desarrollo Regional (Universidad de Castilla La-Mancha)
|
| 34 |
* Campus Universitario s/n
|
| 35 |
* 02071 Alabacete
|
| 36 |
* Spain
|
| 37 |
*
|
| 38 |
* +34 967 599 200
|
| 39 |
*/
|
| 40 |
|
| 41 |
package org.gvsig.remotesensing.classification; |
| 42 |
|
| 43 |
import java.util.ArrayList; |
| 44 |
|
| 45 |
import org.gvsig.andami.PluginServices; |
| 46 |
import org.gvsig.app.project.documents.view.gui.DefaultViewPanel; |
| 47 |
import org.gvsig.fmap.raster.layers.FLyrRasterSE; |
| 48 |
import org.gvsig.raster.buffer.RasterBuffer; |
| 49 |
import org.gvsig.raster.dataset.IBuffer; |
| 50 |
import org.gvsig.raster.grid.GridException; |
| 51 |
import org.gvsig.raster.grid.roi.ROI; |
| 52 |
import org.gvsig.raster.util.RasterToolsUtil; |
| 53 |
|
| 54 |
import Jama.Matrix; |
| 55 |
|
| 56 |
|
| 57 |
/** ClassificationMaximumLikelihoodProcess implementa el m?todo de clasificaci?n de
|
| 58 |
* m?xima probabilidad.
|
| 59 |
*
|
| 60 |
* @see ClassificationGeneralProcess
|
| 61 |
* @author Alejandro Mu?oz Sanchez (alejandro.munoz@uclm.es)
|
| 62 |
* @author Diego Guerrero Sevilla (diego.guerrero@uclm.es)
|
| 63 |
* @version 19/10/2007
|
| 64 |
*/
|
| 65 |
|
| 66 |
public class ClassificationMaximumLikelihoodProcess extends ClassificationGeneralProcess{ |
| 67 |
|
| 68 |
private Matrix Y = null; |
| 69 |
private Matrix result = null; |
| 70 |
private Matrix inverseVarCovMatrix [] = null; |
| 71 |
private double detS [] = null; |
| 72 |
private double means[][] = null; |
| 73 |
private int bandCount = 0; |
| 74 |
|
| 75 |
/**
|
| 76 |
* Class constructor.
|
| 77 |
*/
|
| 78 |
public ClassificationMaximumLikelihoodProcess(){
|
| 79 |
|
| 80 |
} |
| 81 |
|
| 82 |
|
| 83 |
/**
|
| 84 |
* Metodo que implementa el clasificador de maxima probabilidad. Para cada
|
| 85 |
* pixel, obtiene la clase que minimiza la expresion: -Ln(P(x))= Ln(|Si|)+Y'*inverse(Si)*Y
|
| 86 |
*
|
| 87 |
* @param array de tipo byte con valores del pixel en cada una de las bandas
|
| 88 |
* @return clase a la que pertenece el pixel
|
| 89 |
*/
|
| 90 |
public int getPixelClassForTypeByte(byte pixelBandsValues[]){ |
| 91 |
double probability=0.0, finalProbability=0.0; |
| 92 |
int claseFinal=0; |
| 93 |
for (int clase=0; clase<numClases;clase++) |
| 94 |
{
|
| 95 |
double[][] y = new double[bandCount][1]; |
| 96 |
for (int i=0;i<bandCount;i++){ |
| 97 |
y[i][0]=pixelBandsValues[i]-means[clase][i];
|
| 98 |
} |
| 99 |
Y = new Matrix(y);
|
| 100 |
result= (Y.transpose().times(inverseVarCovMatrix[clase])).times(Y); |
| 101 |
// Obtencion probabilidad de pertenencia del pixel a la clase clase
|
| 102 |
probability= Math.log(detS[clase])+ result.get(0, 0); |
| 103 |
if(clase==0) |
| 104 |
finalProbability=probability; |
| 105 |
else if(probability<finalProbability){ |
| 106 |
finalProbability=probability; |
| 107 |
claseFinal=clase; |
| 108 |
} |
| 109 |
} |
| 110 |
return claseFinal;
|
| 111 |
} |
| 112 |
|
| 113 |
|
| 114 |
/**
|
| 115 |
* Metodo que implementa el clasificador de maxima probabilidad.
|
| 116 |
* Para cada pixel, obtiene la calase que minimiza la expresion: -Ln(P(x))= Ln(|Si|)+Y'* inverse(Si)*Y
|
| 117 |
*
|
| 118 |
* @param array de tipo short con valores del pixel en cada una de las bandas
|
| 119 |
* @return clase a la que pertenece el pixel (por el metodo de maxima probabilidad)
|
| 120 |
*/
|
| 121 |
public int getPixelClassForTypeShort(short pixelBandsValues[]){ |
| 122 |
double probability=0.0, finalProbability=0.0; |
| 123 |
int claseFinal=0; |
| 124 |
for (int clase=0; clase<numClases;clase++) |
| 125 |
{
|
| 126 |
double[][] y = new double[bandCount][1]; |
| 127 |
for (int i=0;i<bandCount;i++){ |
| 128 |
y[i][0]=pixelBandsValues[i]-means[clase][i];
|
| 129 |
} |
| 130 |
Y = new Matrix(y);
|
| 131 |
result= (Y.transpose().times(inverseVarCovMatrix[clase])).times(Y); |
| 132 |
// Obtencion probabilidad de pertenencia del pixel a la clase clase
|
| 133 |
probability= Math.log(detS[clase])+ result.get(0, 0); |
| 134 |
if(clase==0) |
| 135 |
finalProbability=probability; |
| 136 |
else if(probability<finalProbability){ |
| 137 |
finalProbability=probability; |
| 138 |
claseFinal=clase; |
| 139 |
} |
| 140 |
} |
| 141 |
return claseFinal;
|
| 142 |
} |
| 143 |
|
| 144 |
|
| 145 |
/**
|
| 146 |
* Metodo que implementa el clasificador de maxima probabilidad.
|
| 147 |
* Para cada pixel, obtiene la calase que minimiza la expresion: -Ln(P(x))= Ln(|Si|)+Y'* inverse(Si)*Y
|
| 148 |
*
|
| 149 |
* @param array de tipo int con valores del pixel en cada una de las bandas
|
| 150 |
* @return clase a la que pertenece el pixel (por el metodo de maxima probabilidad)
|
| 151 |
*/
|
| 152 |
public int getPixelClassForTypeInt(int pixelBandsValues[]){ |
| 153 |
double probability=0.0, finalProbability=0.0; |
| 154 |
int claseFinal=0; |
| 155 |
for (int clase=0; clase<numClases;clase++) |
| 156 |
{
|
| 157 |
double[][] y = new double[bandCount][1]; |
| 158 |
for (int i=0;i<bandCount;i++){ |
| 159 |
y[i][0]=pixelBandsValues[i]-means[clase][i];
|
| 160 |
} |
| 161 |
Y = new Matrix(y);
|
| 162 |
result= (Y.transpose().times(inverseVarCovMatrix[clase])).times(Y); |
| 163 |
// Obtencion probabilidad de pertenencia del pixel a la clase clase
|
| 164 |
probability= Math.log(detS[clase])+ result.get(0, 0); |
| 165 |
if(clase==0) |
| 166 |
finalProbability=probability; |
| 167 |
else if(probability<finalProbability){ |
| 168 |
finalProbability=probability; |
| 169 |
claseFinal=clase; |
| 170 |
} |
| 171 |
} |
| 172 |
return claseFinal;
|
| 173 |
} |
| 174 |
|
| 175 |
|
| 176 |
/**
|
| 177 |
* Metodo que implementa el clasificador de maxima probabilidad.
|
| 178 |
* Para cada pixel, obtiene la calase que minimiza la expresion: -Ln(P(x))= Ln(|Si|)+Y'* inverse(Si)*Y
|
| 179 |
*
|
| 180 |
* @param array de tipo float con valores del pixel en cada una de las bandas
|
| 181 |
* @return clase a la que pertenece el pixel (por el metodo de maxima probabilidad)
|
| 182 |
*/
|
| 183 |
public int getPixelClassForTypeFloat(float pixelBandsValues[]){ |
| 184 |
double probability=0.0, finalProbability=0.0; |
| 185 |
int claseFinal=0; |
| 186 |
for (int clase=0; clase<numClases;clase++) |
| 187 |
{
|
| 188 |
double[][] y = new double[bandCount][1]; |
| 189 |
for (int i=0;i<bandCount;i++){ |
| 190 |
y[i][0]=pixelBandsValues[i]-means[clase][i];
|
| 191 |
} |
| 192 |
Y = new Matrix(y);
|
| 193 |
result= (Y.transpose().times(inverseVarCovMatrix[clase])).times(Y); |
| 194 |
// Obtencion probabilidad de pertenencia del pixel a la clase clase
|
| 195 |
probability= Math.log(detS[clase])+ result.get(0, 0); |
| 196 |
if(clase==0) |
| 197 |
finalProbability=probability; |
| 198 |
else if(probability<finalProbability){ |
| 199 |
finalProbability=probability; |
| 200 |
claseFinal=clase; |
| 201 |
} |
| 202 |
} |
| 203 |
return claseFinal;
|
| 204 |
} |
| 205 |
|
| 206 |
|
| 207 |
/**
|
| 208 |
* Metodo que implementa el clasificador de maxima probabilidad.
|
| 209 |
* Para cada pixel, obtiene la calase que minimiza la expresion: -Ln(P(x))= Ln(|Si|)+Y'* inverse(Si)*Y
|
| 210 |
*
|
| 211 |
* @param array de tipo double con valores del pixel en cada una de las bandas
|
| 212 |
* @return clase a la que pertenece el pixel (por el metodo de maxima probabilidad)
|
| 213 |
*/
|
| 214 |
public int getPixelClassForTypeDouble(double pixelBandsValues[]){ |
| 215 |
double probability=0.0, finalProbability=0.0; |
| 216 |
int claseFinal=0; |
| 217 |
for (int clase=0; clase<numClases;clase++) |
| 218 |
{
|
| 219 |
double[][] y = new double[bandCount][1]; |
| 220 |
for (int i=0;i<bandCount;i++){ |
| 221 |
y[i][0]=pixelBandsValues[i]-means[clase][i];
|
| 222 |
} |
| 223 |
Y = new Matrix(y);
|
| 224 |
result= (Y.transpose().times(inverseVarCovMatrix[clase])).times(Y); |
| 225 |
// Obtencion probabilidad de pertenencia del pixel a la clase clase
|
| 226 |
probability= Math.log(detS[clase])+ result.get(0, 0); |
| 227 |
if(clase==0) |
| 228 |
finalProbability=probability; |
| 229 |
else if(probability<finalProbability){ |
| 230 |
finalProbability=probability; |
| 231 |
claseFinal=clase; |
| 232 |
} |
| 233 |
} |
| 234 |
return claseFinal;
|
| 235 |
} |
| 236 |
|
| 237 |
/** Metodo que recoge los parametros del proceso de clasificacion de
|
| 238 |
* m?xima probabilidad
|
| 239 |
* <LI>rasterSE: Capa de entrada para la clasificaci?n</LI>
|
| 240 |
* <LI> rois: lista de rois</LI>
|
| 241 |
* <LI> bandList:bandas habilitadas </LI>
|
| 242 |
* <LI>view: vista sobre la que se carga la capa al acabar el proceso</LI>
|
| 243 |
* <LI>filename: path con el fichero de salida</LI>
|
| 244 |
*/
|
| 245 |
public void init() { |
| 246 |
rasterSE= (FLyrRasterSE)getParam("layer");
|
| 247 |
rois = (ArrayList)getParam("rois"); |
| 248 |
view=(DefaultViewPanel)getParam("view");
|
| 249 |
filename= getStringParam("filename");
|
| 250 |
bandList = (int[])getParam("bandList"); |
| 251 |
numClases = rois.size(); |
| 252 |
} |
| 253 |
|
| 254 |
|
| 255 |
/** Proceso de clasificaci?n de m?xima probabilidad */
|
| 256 |
public void process() throws InterruptedException { |
| 257 |
|
| 258 |
setGrid(); |
| 259 |
rasterResult= RasterBuffer.getBuffer(IBuffer.TYPE_BYTE, inputGrid.getRasterBuf().getWidth(), inputGrid.getRasterBuf().getHeight(), 1, true); |
| 260 |
int c=0; |
| 261 |
int iNY= inputGrid.getLayerNY();
|
| 262 |
int iNX= inputGrid.getLayerNX();
|
| 263 |
bandCount = inputGrid.getBandCount(); |
| 264 |
|
| 265 |
means = new double[numClases][bandCount]; |
| 266 |
for (int clase=0; clase<numClases; clase++) |
| 267 |
for (int i=0;i<bandCount;i++){ |
| 268 |
((ROI)rois.get(clase)).setBandToOperate(bandList[i]); |
| 269 |
try{
|
| 270 |
means[clase][i]=((ROI)rois.get(clase)).getMeanValue(); |
| 271 |
}catch (GridException e) {
|
| 272 |
RasterToolsUtil.messageBoxError(PluginServices.getText(this, "grid_error"), this, e); |
| 273 |
} |
| 274 |
} |
| 275 |
|
| 276 |
|
| 277 |
|
| 278 |
int inputGridNX = inputGrid.getNX();
|
| 279 |
int datType = inputGrid.getRasterBuf().getDataType();
|
| 280 |
|
| 281 |
// Se calculan las inversas de las matrices de Varianza-covarianza de todas las rois y se almacenan en inverseVarCovMAtrix
|
| 282 |
Matrix S = null;
|
| 283 |
Matrix inverseS = null;
|
| 284 |
inverseVarCovMatrix= new Matrix[numClases];
|
| 285 |
detS = new double [numClases]; |
| 286 |
double varCovarMatrix[][] = null; |
| 287 |
double subMatrix[][] = null; |
| 288 |
|
| 289 |
for (int i=0; i<numClases;i++){ |
| 290 |
try{
|
| 291 |
varCovarMatrix = ((ROI)rois.get(i)).getVarCovMatrix(); |
| 292 |
}catch (GridException e) {
|
| 293 |
RasterToolsUtil.messageBoxError(PluginServices.getText(this, "grid_error"), this, e); |
| 294 |
} |
| 295 |
if (bandList.length != rasterSE.getBandCount()){
|
| 296 |
/*
|
| 297 |
* Extraer la submatiz correspondiente a las bandas que intervienen:
|
| 298 |
*/
|
| 299 |
subMatrix = new double[bandList.length][bandList.length]; |
| 300 |
for (int iBand = 0; iBand < bandList.length; iBand++) |
| 301 |
for (int jBand = 0; jBand < bandList.length; jBand++) |
| 302 |
subMatrix[iBand][jBand]=varCovarMatrix[bandList[iBand]][bandList[jBand]]; |
| 303 |
S = new Matrix(subMatrix);
|
| 304 |
inverseS = S.inverse(); |
| 305 |
detS[i]=S.det(); |
| 306 |
}else
|
| 307 |
try {
|
| 308 |
S = new Matrix(((ROI)rois.get(i)).getVarCovMatrix());
|
| 309 |
inverseS = S.inverse(); |
| 310 |
detS[i] = S.det(); |
| 311 |
|
| 312 |
} catch (RuntimeException e) { |
| 313 |
RasterToolsUtil.messageBoxError(PluginServices.getText(this, "error_clasificacion_roi") +((ROI)rois.get(i)).getName(),this); |
| 314 |
return;
|
| 315 |
} catch (GridException e) {
|
| 316 |
RasterToolsUtil.messageBoxError(PluginServices.getText(this, "grid_error"), this, e); |
| 317 |
} |
| 318 |
|
| 319 |
inverseVarCovMatrix[i]= inverseS; |
| 320 |
} |
| 321 |
|
| 322 |
// Caso Buffer tipo Byte
|
| 323 |
if (datType == RasterBuffer.TYPE_BYTE){
|
| 324 |
|
| 325 |
byte data[]= new byte[inputGrid.getBandCount()]; |
| 326 |
for(int i=0; i<iNY;i++){ |
| 327 |
for(int j=0; j<iNX;j++){ |
| 328 |
inputGrid.getRasterBuf().getElemByte(i, j, data); |
| 329 |
c= getPixelClassForTypeByte(data); |
| 330 |
rasterResult.setElem(i, j, 0,(byte) c); |
| 331 |
} |
| 332 |
percent = i*100/inputGridNX;
|
| 333 |
} |
| 334 |
} |
| 335 |
|
| 336 |
// Caso Buffer tipo Short
|
| 337 |
if (datType == RasterBuffer.TYPE_SHORT){
|
| 338 |
short data[]= new short[inputGrid.getBandCount()]; |
| 339 |
for(int i=0; i<iNY;i++){ |
| 340 |
for(int j=0; j<iNX;j++){ |
| 341 |
inputGrid.getRasterBuf().getElemShort(i, j, data); |
| 342 |
c= getPixelClassForTypeShort(data); |
| 343 |
rasterResult.setElem(i, j, 0,(byte)c); |
| 344 |
} |
| 345 |
percent = i*100/inputGridNX;
|
| 346 |
} |
| 347 |
} |
| 348 |
|
| 349 |
// Caso Buffer tipo Int
|
| 350 |
if (datType == RasterBuffer.TYPE_INT){
|
| 351 |
int data[]= new int[inputGrid.getBandCount()]; |
| 352 |
for(int i=0; i<iNY;i++){ |
| 353 |
for(int j=0; j<iNX;j++){ |
| 354 |
inputGrid.getRasterBuf().getElemInt(i, j, data); |
| 355 |
c= getPixelClassForTypeInt(data); |
| 356 |
rasterResult.setElem(i, j, 0,(byte) c); |
| 357 |
} |
| 358 |
percent = i*100/inputGridNX;
|
| 359 |
} |
| 360 |
} |
| 361 |
|
| 362 |
|
| 363 |
// Caso Buffer tipo Float
|
| 364 |
if (datType == RasterBuffer.TYPE_FLOAT){
|
| 365 |
float data[]= new float[inputGrid.getBandCount()]; |
| 366 |
for(int i=0; i<iNY;i++){ |
| 367 |
for(int j=0; j<iNX;j++){ |
| 368 |
inputGrid.getRasterBuf().getElemFloat(i, j, data); |
| 369 |
c= getPixelClassForTypeFloat(data); |
| 370 |
rasterResult.setElem(i, j, 0,(byte) c); |
| 371 |
} |
| 372 |
percent = i*100/inputGridNX;
|
| 373 |
} |
| 374 |
} |
| 375 |
|
| 376 |
|
| 377 |
// Caso Buffer tipo Double
|
| 378 |
if (datType == RasterBuffer.TYPE_DOUBLE){
|
| 379 |
double data[]= new double[inputGrid.getBandCount()]; |
| 380 |
for(int i=0; i<iNY;i++){ |
| 381 |
for(int j=0; j<iNX;j++){ |
| 382 |
inputGrid.getRasterBuf().getElemDouble(i, j, data); |
| 383 |
c= getPixelClassForTypeDouble(data); |
| 384 |
rasterResult.setElem(i, j, 0,(byte) c); |
| 385 |
} |
| 386 |
percent = i*100/inputGridNX;
|
| 387 |
} |
| 388 |
} |
| 389 |
|
| 390 |
writeToFile(); |
| 391 |
// incrementableTask.processFinalize();
|
| 392 |
} |
| 393 |
} |