root / org.gvsig.toolbox / trunk / org.gvsig.toolbox / org.gvsig.toolbox.algorithm / src / main / java / es / unex / sextante / gridAnalysis / supervisedClassificationB / SupervisedClassificationBAlgorithm.java @ 59
History | View | Annotate | Download (11.3 KB)
1 |
package es.unex.sextante.gridAnalysis.supervisedClassificationB; |
---|---|
2 |
|
3 |
import java.util.ArrayList; |
4 |
import java.util.HashMap; |
5 |
import java.util.Iterator; |
6 |
import java.util.Set; |
7 |
|
8 |
import es.unex.sextante.additionalInfo.AdditionalInfoMultipleInput; |
9 |
import es.unex.sextante.core.AnalysisExtent; |
10 |
import es.unex.sextante.core.GeoAlgorithm; |
11 |
import es.unex.sextante.core.Sextante; |
12 |
import es.unex.sextante.dataObjects.IRasterLayer; |
13 |
import es.unex.sextante.dataObjects.IRecord; |
14 |
import es.unex.sextante.dataObjects.IRecordsetIterator; |
15 |
import es.unex.sextante.dataObjects.ITable; |
16 |
import es.unex.sextante.exceptions.GeoAlgorithmExecutionException; |
17 |
import es.unex.sextante.exceptions.RepeatedParameterNameException; |
18 |
import es.unex.sextante.parameters.RasterLayerAndBand; |
19 |
|
20 |
public class SupervisedClassificationBAlgorithm |
21 |
extends
|
22 |
GeoAlgorithm { |
23 |
|
24 |
public static final String INPUT = "INPUT"; |
25 |
public static final String METHOD = "METHOD"; |
26 |
public static final String CLASSIFICATION = "CLASSIFICATION"; |
27 |
public static final String CLASSES = "CLASSES"; |
28 |
public static final String TABLE = "TABLE"; |
29 |
|
30 |
public static final int METHOD_PARALELLPIPED = 0; |
31 |
public static final int METHOD_MIN_DISTANCE = 1; |
32 |
public static final int METHOD_MAX_LIKELIHOOD = 2; |
33 |
|
34 |
private IRasterLayer[] m_Window; |
35 |
private IRasterLayer m_Output;
|
36 |
private ArrayList m_Bands; |
37 |
private HashMap m_Classes; |
38 |
private int[] m_iBands; |
39 |
|
40 |
|
41 |
@Override
|
42 |
public void defineCharacteristics() { |
43 |
|
44 |
final String sMethod[] = { Sextante.getText("Parallelepiped"), Sextante.getText("Minimum_distance"), |
45 |
Sextante.getText("Maximum_likelihood") };
|
46 |
|
47 |
setName(Sextante.getText("Supervised_classification") + "(B)"); |
48 |
setGroup(Sextante.getText("Raster_layer_analysis"));
|
49 |
setUserCanDefineAnalysisExtent(true);
|
50 |
|
51 |
try {
|
52 |
m_Parameters.addMultipleInput(INPUT, Sextante.getText("Bands"), AdditionalInfoMultipleInput.DATA_TYPE_BAND, true); |
53 |
m_Parameters.addInputTable(TABLE, Sextante.getText("Classes"), true); |
54 |
m_Parameters.addSelection(METHOD, Sextante.getText("Method"), sMethod);
|
55 |
addOutputRasterLayer(CLASSIFICATION, Sextante.getText("Classification"));
|
56 |
//addOutputTable(CLASSES, Sextante.getText("Classes"));
|
57 |
} |
58 |
catch (final RepeatedParameterNameException e) { |
59 |
Sextante.addErrorToLog(e); |
60 |
} |
61 |
|
62 |
} |
63 |
|
64 |
|
65 |
@Override
|
66 |
public boolean processAlgorithm() throws GeoAlgorithmExecutionException { |
67 |
|
68 |
int i;
|
69 |
AnalysisExtent ge; |
70 |
|
71 |
final int iMethod = m_Parameters.getParameterValueAsInt(METHOD); |
72 |
m_Bands = m_Parameters.getParameterValueAsArrayList(INPUT); |
73 |
|
74 |
|
75 |
if (m_Bands.size() == 0) { |
76 |
return false; |
77 |
} |
78 |
|
79 |
m_Classes = new HashMap(); |
80 |
|
81 |
getClassInformation(); |
82 |
|
83 |
if (m_Task.isCanceled()) {
|
84 |
return false; |
85 |
} |
86 |
|
87 |
m_Output = getNewRasterLayer(CLASSIFICATION, Sextante.getText("Classification"), IRasterLayer.RASTER_DATA_TYPE_SHORT);
|
88 |
m_Output.setNoDataValue(-1);
|
89 |
ge = m_Output.getWindowGridExtent(); |
90 |
|
91 |
m_Window = new IRasterLayer[m_Bands.size()];
|
92 |
m_iBands = new int[m_Bands.size()]; |
93 |
for (i = 0; i < m_Window.length; i++) { |
94 |
final RasterLayerAndBand band = (RasterLayerAndBand) m_Bands.get(i);
|
95 |
m_iBands[i] = band.getBand(); |
96 |
m_Window[i] = band.getRasterLayer(); |
97 |
m_Window[i].setWindowExtent(ge); |
98 |
} |
99 |
|
100 |
switch (iMethod) {
|
101 |
case 0: |
102 |
doParalellpiped(); |
103 |
case 1: |
104 |
default:
|
105 |
doMinimumDistance(); |
106 |
case 2: |
107 |
doMaximumLikelihood(); |
108 |
} |
109 |
|
110 |
return !m_Task.isCanceled();
|
111 |
|
112 |
} |
113 |
|
114 |
|
115 |
private void getClassInformation() throws GeoAlgorithmExecutionException { |
116 |
|
117 |
try {
|
118 |
final ITable table = m_Parameters.getParameterValueAsTable(TABLE);
|
119 |
m_Window = new IRasterLayer[m_Bands.size()];
|
120 |
|
121 |
final IRecordsetIterator iter = table.iterator();
|
122 |
while (iter.hasNext()) {
|
123 |
final IRecord record = iter.next();
|
124 |
final String sClassName = record.getValue(0).toString(); |
125 |
final ArrayList stats = new ArrayList(); |
126 |
for (int i = 0; i < m_Window.length; i++) { |
127 |
final String sFieldName = m_Window[i].getName() + "|" + Integer.toString(m_iBands[i] + 1); |
128 |
final MeanAndStdDev masd = new MeanAndStdDev(); |
129 |
boolean bMatchFound = false; |
130 |
for (int j = 1; j < table.getFieldCount(); j += 2) { |
131 |
if (table.getFieldName(j).equals(sFieldName)) {
|
132 |
masd.mean = Double.parseDouble(record.getValue(j).toString());
|
133 |
masd.stdDev = Double.parseDouble(record.getValue(j + 1).toString()); |
134 |
bMatchFound = true;
|
135 |
} |
136 |
} |
137 |
if (!bMatchFound) {
|
138 |
throw new GeoAlgorithmExecutionException(Sextante.getText("Error_reading_table")); |
139 |
} |
140 |
stats.add(masd); |
141 |
} |
142 |
m_Classes.put(sClassName, stats); |
143 |
} |
144 |
} |
145 |
catch (final Exception e) { |
146 |
throw new GeoAlgorithmExecutionException(Sextante.getText("Error_reading_table")); |
147 |
} |
148 |
|
149 |
|
150 |
} |
151 |
|
152 |
|
153 |
private void doParalellpiped() { |
154 |
|
155 |
int iNX, iNY;
|
156 |
int x, y;
|
157 |
int iMatchingClass = 0; |
158 |
int iClass, iGrid;
|
159 |
final double dMean[][] = new double[m_Classes.size()][m_Window.length]; |
160 |
final double dStdDev[][] = new double[m_Classes.size()][m_Window.length]; |
161 |
double dValue;
|
162 |
ArrayList stats;
|
163 |
MeanAndStdDev substats; |
164 |
Set set;
|
165 |
Iterator iter;
|
166 |
|
167 |
iNX = m_Output.getWindowGridExtent().getNX(); |
168 |
iNY = m_Output.getWindowGridExtent().getNY(); |
169 |
|
170 |
set = m_Classes.keySet(); |
171 |
iter = set.iterator(); |
172 |
iClass = 0;
|
173 |
while (iter.hasNext()) {
|
174 |
stats = (ArrayList) m_Classes.get(iter.next());
|
175 |
for (iGrid = 0; iGrid < m_Window.length; iGrid++) { |
176 |
substats = ((MeanAndStdDev) stats.get(iGrid)); |
177 |
dMean[iClass][iGrid] = substats.mean; |
178 |
dStdDev[iClass][iGrid] = substats.stdDev; |
179 |
} |
180 |
iClass++; |
181 |
} |
182 |
|
183 |
for (y = 0; y < iNY; y++) { |
184 |
for (x = 0; x < iNX; x++) { |
185 |
for (iClass = 0; iClass < m_Classes.size(); iClass++) { |
186 |
iMatchingClass = iClass; |
187 |
for (iGrid = 0; iGrid < m_Window.length; iGrid++) { |
188 |
dValue = m_Window[iGrid].getCellValueAsDouble(x, y); |
189 |
if (!m_Window[iGrid].isNoDataValue(dValue)) {
|
190 |
if (Math.abs(m_Window[iGrid].getCellValueAsDouble(x, y) - dMean[iClass][iGrid]) > dStdDev[iClass][iGrid]) { |
191 |
iMatchingClass = -1;
|
192 |
break;
|
193 |
} |
194 |
} |
195 |
else {
|
196 |
break;
|
197 |
} |
198 |
} |
199 |
if (iMatchingClass != -1) { |
200 |
break;
|
201 |
} |
202 |
} |
203 |
if (iMatchingClass != -1) { |
204 |
m_Output.setCellValue(x, y, iMatchingClass + 1);
|
205 |
} |
206 |
else {
|
207 |
m_Output.setNoData(x, y); |
208 |
} |
209 |
} |
210 |
} |
211 |
|
212 |
} |
213 |
|
214 |
|
215 |
private void doMinimumDistance() { |
216 |
|
217 |
int iNX, iNY;
|
218 |
int x, y;
|
219 |
int iClass, iGrid, iMin = 0; |
220 |
final double dMean[][] = new double[m_Classes.size()][m_Window.length]; |
221 |
double dMin, d, e;
|
222 |
double dValue;
|
223 |
ArrayList stats;
|
224 |
Set set;
|
225 |
Iterator iter;
|
226 |
|
227 |
iNX = m_Output.getWindowGridExtent().getNX(); |
228 |
iNY = m_Output.getWindowGridExtent().getNY(); |
229 |
|
230 |
set = m_Classes.keySet(); |
231 |
iter = set.iterator(); |
232 |
iClass = 0;
|
233 |
while (iter.hasNext()) {
|
234 |
stats = (ArrayList) m_Classes.get(iter.next());
|
235 |
for (iGrid = 0; iGrid < m_Window.length; iGrid++) { |
236 |
dMean[iClass][iGrid] = ((MeanAndStdDev) stats.get(iGrid)).mean; |
237 |
} |
238 |
iClass++; |
239 |
} |
240 |
|
241 |
for (y = 0; y < iNY; y++) { |
242 |
for (x = 0; x < iNX; x++) { |
243 |
for (iClass = 0, dMin = -1.0; iClass < m_Classes.size(); iClass++) { |
244 |
for (iGrid = 0, d = 0.0; iGrid < m_Window.length; iGrid++) { |
245 |
dValue = m_Window[iGrid].getCellValueAsDouble(x, y); |
246 |
if (!m_Window[iGrid].isNoDataValue(dValue)) {
|
247 |
e = m_Window[iGrid].getCellValueAsDouble(x, y) - dMean[iClass][iGrid]; |
248 |
d += e * e; |
249 |
if ((dMin < 0.0) || (dMin > d)) { |
250 |
dMin = d; |
251 |
iMin = iClass; |
252 |
} |
253 |
} |
254 |
else {
|
255 |
dMin = -1;
|
256 |
} |
257 |
} |
258 |
} |
259 |
|
260 |
if (dMin >= 0.0) { |
261 |
m_Output.setCellValue(x, y, iMin + 1);
|
262 |
} |
263 |
else {
|
264 |
m_Output.setNoData(x, y); |
265 |
} |
266 |
} |
267 |
} |
268 |
|
269 |
} |
270 |
|
271 |
|
272 |
private void doMaximumLikelihood() { |
273 |
|
274 |
int iNX, iNY;
|
275 |
int x, y;
|
276 |
int iClass, iGrid, iMax = 0; |
277 |
final double dMean[][] = new double[m_Classes.size()][m_Window.length]; |
278 |
final double dStdDev[][] = new double[m_Classes.size()][m_Window.length]; |
279 |
final double dK[][] = new double[m_Classes.size()][m_Window.length]; |
280 |
double dMax, d, e;
|
281 |
double dValue;
|
282 |
ArrayList stats;
|
283 |
MeanAndStdDev substats; |
284 |
Set set;
|
285 |
Iterator iter;
|
286 |
|
287 |
iNX = m_Output.getWindowGridExtent().getNX(); |
288 |
iNY = m_Output.getWindowGridExtent().getNY(); |
289 |
|
290 |
set = m_Classes.keySet(); |
291 |
iter = set.iterator(); |
292 |
iClass = 0;
|
293 |
while (iter.hasNext()) {
|
294 |
stats = (ArrayList) m_Classes.get(iter.next());
|
295 |
for (iGrid = 0; iGrid < m_Window.length; iGrid++) { |
296 |
substats = ((MeanAndStdDev) stats.get(iGrid)); |
297 |
dMean[iClass][iGrid] = substats.mean; |
298 |
dStdDev[iClass][iGrid] = substats.stdDev; |
299 |
dK[iClass][iGrid] = 1.0 / (dStdDev[iClass][iGrid] * Math.sqrt(2.0 * Math.PI)); |
300 |
} |
301 |
iClass++; |
302 |
} |
303 |
|
304 |
for (y = 0; y < iNY; y++) { |
305 |
for (x = 0; x < iNX; x++) { |
306 |
for (iClass = 0, dMax = 0.0; iClass < m_Classes.size(); iClass++) { |
307 |
for (iGrid = 0, d = 0.0; iGrid < m_Window.length; iGrid++) { |
308 |
dValue = m_Window[iGrid].getCellValueAsDouble(x, y); |
309 |
if (!m_Window[iGrid].isNoDataValue(dValue)) {
|
310 |
e = (m_Window[iGrid].getCellValueAsDouble(x, y) - dMean[iClass][iGrid]) / dStdDev[iClass][iGrid]; |
311 |
e = dK[iClass][iGrid] * Math.exp(-0.5 * e * e); |
312 |
d += e * e; |
313 |
if (dMax < d) {
|
314 |
dMax = d; |
315 |
iMax = iClass; |
316 |
} |
317 |
} |
318 |
else {
|
319 |
dMax = -1;
|
320 |
} |
321 |
} |
322 |
} |
323 |
|
324 |
if (dMax > 0.0) { |
325 |
m_Output.setCellValue(x, y, iMax + 1);
|
326 |
} |
327 |
else {
|
328 |
m_Output.setNoData(x, y); |
329 |
} |
330 |
} |
331 |
} |
332 |
|
333 |
} |
334 |
|
335 |
private class MeanAndStdDev { |
336 |
|
337 |
public double mean = 0; |
338 |
public double stdDev = 0; |
339 |
|
340 |
} |
341 |
|
342 |
} |