root / org.gvsig.toolbox / trunk / org.gvsig.toolbox / org.gvsig.toolbox.algorithm / src / main / java / es / unex / sextante / statisticalMethods / pca / PCAAlgorithm.java @ 59
History | View | Annotate | Download (6.97 KB)
1 |
|
---|---|
2 |
|
3 |
package es.unex.sextante.statisticalMethods.pca; |
4 |
|
5 |
|
6 |
import java.util.ArrayList; |
7 |
import java.util.Arrays; |
8 |
|
9 |
import Jama.EigenvalueDecomposition; |
10 |
import Jama.Matrix; |
11 |
import es.unex.sextante.additionalInfo.AdditionalInfoMultipleInput; |
12 |
import es.unex.sextante.additionalInfo.AdditionalInfoNumericalValue; |
13 |
import es.unex.sextante.core.GeoAlgorithm; |
14 |
import es.unex.sextante.core.Sextante; |
15 |
import es.unex.sextante.dataObjects.IRasterLayer; |
16 |
import es.unex.sextante.exceptions.GeoAlgorithmExecutionException; |
17 |
import es.unex.sextante.exceptions.RepeatedParameterNameException; |
18 |
import es.unex.sextante.outputs.OutputRasterLayer; |
19 |
import es.unex.sextante.parameters.RasterLayerAndBand; |
20 |
|
21 |
|
22 |
public class PCAAlgorithm |
23 |
extends
|
24 |
GeoAlgorithm { |
25 |
|
26 |
public static final String RESULT = "RESULT"; |
27 |
public static final String NBANDS = "NBANDS"; |
28 |
public static final String INPUT = "INPUT"; |
29 |
|
30 |
private static double NODATA = -9999999.; |
31 |
|
32 |
private ArrayList m_RasterLayers; |
33 |
private IRasterLayer[] m_Windows; |
34 |
private int m_iBands[]; |
35 |
private int m_iNX, m_iNY; |
36 |
private double m_dMean[]; |
37 |
private int m_iNumBands; |
38 |
|
39 |
|
40 |
@Override
|
41 |
public void defineCharacteristics() { |
42 |
|
43 |
setUserCanDefineAnalysisExtent(true);
|
44 |
setGroup(Sextante.getText("Statistical_methods"));
|
45 |
setName(Sextante.getText("Principal_Components_Analysis"));
|
46 |
|
47 |
try {
|
48 |
m_Parameters.addMultipleInput(INPUT, Sextante.getText("Input_bands"), AdditionalInfoMultipleInput.DATA_TYPE_BAND, true); |
49 |
m_Parameters.addNumericalValue(NBANDS, Sextante.getText("Bands_number"),
|
50 |
AdditionalInfoNumericalValue.NUMERICAL_VALUE_INTEGER, 1, 1, Integer.MAX_VALUE); |
51 |
addOutputRasterLayer(RESULT, Sextante.getText("Principal_Components_Analysis"),
|
52 |
OutputRasterLayer.NUMBER_OF_BANDS_UNDEFINED); |
53 |
} |
54 |
catch (final RepeatedParameterNameException e) { |
55 |
Sextante.addErrorToLog(e); |
56 |
} |
57 |
|
58 |
} |
59 |
|
60 |
|
61 |
@Override
|
62 |
public boolean processAlgorithm() throws GeoAlgorithmExecutionException { |
63 |
|
64 |
int i;
|
65 |
int x, y;
|
66 |
int iBand;
|
67 |
int iInputBand;
|
68 |
double dValue;
|
69 |
double dPCValue;
|
70 |
|
71 |
m_RasterLayers = m_Parameters.getParameterValueAsArrayList(INPUT); |
72 |
m_iNumBands = m_Parameters.getParameterValueAsInt(NBANDS); |
73 |
|
74 |
if ((m_RasterLayers.size() == 0) || (m_iNumBands > m_RasterLayers.size())) { |
75 |
throw new GeoAlgorithmExecutionException(Sextante.getText("Error_not_enough_bands_selected")); |
76 |
} |
77 |
|
78 |
m_iNX = getAnalysisExtent().getNX(); |
79 |
m_iNY = getAnalysisExtent().getNY(); |
80 |
|
81 |
final double coVar[][] = getCovarMatrix(); |
82 |
final Matrix coVarMatrix = new Matrix(coVar); |
83 |
final EigenvalueDecomposition eigenvalueDecomp = new EigenvalueDecomposition(coVarMatrix); |
84 |
final Matrix eigenvectors = eigenvalueDecomp.getV();
|
85 |
final double[] eigenvalues = eigenvalueDecomp.getRealEigenvalues(); |
86 |
final ValueAndOrder[] vao = new ValueAndOrder[eigenvalues.length]; |
87 |
for (i = 0; i < eigenvalues.length; i++) { |
88 |
vao[i] = new ValueAndOrder(eigenvalues[i], i);
|
89 |
} |
90 |
|
91 |
Arrays.sort(vao);
|
92 |
final IRasterLayer pca = getNewRasterLayer(RESULT, Sextante.getText("Principal_Components_Analysis"), |
93 |
IRasterLayer.RASTER_DATA_TYPE_DOUBLE, m_iNumBands); |
94 |
|
95 |
for (y = 0; y < m_iNY; y++) { |
96 |
for (x = 0; x < m_iNX; x++) { |
97 |
for (iBand = 0; (iBand < m_iNumBands) && setProgress(iBand, m_iNumBands); iBand++) { |
98 |
dPCValue = 0;
|
99 |
for (iInputBand = 0; iInputBand < m_RasterLayers.size(); iInputBand++) { |
100 |
dValue = m_Windows[iInputBand].getCellValueAsDouble(x, y, m_iBands[iInputBand]); |
101 |
dPCValue += (eigenvectors.get(iInputBand, vao[iBand].getOrder()) * dValue); |
102 |
} |
103 |
pca.setCellValue(x, y, iBand, dPCValue); |
104 |
} |
105 |
} |
106 |
} |
107 |
|
108 |
|
109 |
return !m_Task.isCanceled();
|
110 |
|
111 |
} |
112 |
|
113 |
|
114 |
private double[][] getCovarMatrix() { |
115 |
|
116 |
int i, j;
|
117 |
|
118 |
final double dCovar[][] = new double[m_RasterLayers.size()][m_RasterLayers.size()]; |
119 |
m_dMean = new double[m_RasterLayers.size()]; |
120 |
|
121 |
m_Windows = new IRasterLayer[m_RasterLayers.size()];
|
122 |
m_iBands = new int[m_RasterLayers.size()]; |
123 |
|
124 |
for (i = 0; i < m_RasterLayers.size(); i++) { |
125 |
final RasterLayerAndBand rab = (RasterLayerAndBand) m_RasterLayers.get(i);
|
126 |
m_Windows[i] = rab.getRasterLayer(); |
127 |
m_Windows[i].setWindowExtent(this.getAnalysisExtent());
|
128 |
m_Windows[i].setInterpolationMethod(IRasterLayer.INTERPOLATION_BSpline); |
129 |
m_dMean[i] = m_Windows[i].getMeanValue(); |
130 |
m_iBands[i] = rab.getBand(); |
131 |
} |
132 |
|
133 |
|
134 |
final int iTotal = (int) (m_RasterLayers.size() * m_RasterLayers.size() / 2.); |
135 |
int iCount = 0; |
136 |
for (i = 0; (i < m_RasterLayers.size() - 1) && setProgress(iCount, iTotal); i++) { |
137 |
dCovar[i][i] = 1.0;
|
138 |
iCount++; |
139 |
for (j = i + 1; j < m_RasterLayers.size(); j++) { |
140 |
dCovar[i][j] = dCovar[j][i] = getCovar(i, j); |
141 |
iCount++; |
142 |
} |
143 |
} |
144 |
|
145 |
|
146 |
return dCovar;
|
147 |
|
148 |
} |
149 |
|
150 |
|
151 |
private double getCovar(final int i, |
152 |
final int j) { |
153 |
|
154 |
int x, y;
|
155 |
int iValues = 0; |
156 |
double dValuei, dValuej;
|
157 |
double dSum = 0; |
158 |
|
159 |
for (y = 0; y < m_iNY; y++) { |
160 |
for (x = 0; x < m_iNX; x++) { |
161 |
dValuei = m_Windows[i].getCellValueAsDouble(x, y, m_iBands[i]); |
162 |
dValuej = m_Windows[j].getCellValueAsDouble(x, y, m_iBands[j]); |
163 |
if (!m_Windows[i].isNoDataValue(dValuei) && !m_Windows[j].isNoDataValue(dValuej)) {
|
164 |
dSum += (dValuei - m_dMean[i]) * (dValuej - m_dMean[j]); |
165 |
iValues++; |
166 |
} |
167 |
} |
168 |
if (iValues > 1) { |
169 |
return dSum / (iValues - 1); |
170 |
} |
171 |
else {
|
172 |
return NODATA;
|
173 |
} |
174 |
} |
175 |
|
176 |
return NODATA;
|
177 |
|
178 |
} |
179 |
|
180 |
private class ValueAndOrder |
181 |
implements
|
182 |
Comparable {
|
183 |
|
184 |
private final int m_iOrder; |
185 |
private final double m_dValue; |
186 |
|
187 |
|
188 |
public ValueAndOrder(final double dValue, |
189 |
final int iOrder) { |
190 |
|
191 |
m_dValue = dValue; |
192 |
m_iOrder = iOrder; |
193 |
|
194 |
} |
195 |
|
196 |
|
197 |
public int getOrder() { |
198 |
|
199 |
return m_iOrder;
|
200 |
|
201 |
} |
202 |
|
203 |
|
204 |
public double getValue() { |
205 |
|
206 |
return m_dValue;
|
207 |
|
208 |
} |
209 |
|
210 |
|
211 |
public int compareTo(final Object vao) throws ClassCastException { |
212 |
|
213 |
if (!(vao instanceof ValueAndOrder)) { |
214 |
throw new ClassCastException(); |
215 |
} |
216 |
|
217 |
final double dValue = ((ValueAndOrder) vao).getValue(); |
218 |
final double dDif = this.m_dValue - dValue; |
219 |
|
220 |
if (dDif > 0.0) { |
221 |
return 1; |
222 |
} |
223 |
else if (dDif < 0.0) { |
224 |
return -1; |
225 |
} |
226 |
else {
|
227 |
return 0; |
228 |
} |
229 |
|
230 |
} |
231 |
} |
232 |
} |