@Override
public BxDocument segmentDocument(BxDocument document) throws AnalysisException {
Map<BxPage, List<Component>> componentMap = new HashMap<BxPage, List<Component>>();
ExecutorService exec = Executors.newFixedThreadPool(PdfNLMContentExtractor.THREADS_NUMBER);
ArrayList<Callable<NumBxPage>> tasks = new ArrayList<Callable<NumBxPage>>();
for (BxPage page : document.getPages()) {
tasks.add(new ComponentCounter(page));
}
List<Future<NumBxPage>> results;
try {
results = exec.invokeAll(tasks);
exec.shutdown();
for (Future<NumBxPage> result : results) {
NumBxPage p = result.get();
componentMap.put(p.page, p.components);
}
} catch (ExecutionException ex) {
throw new AnalysisException("Cannot segment pages!", ex);
} catch (InterruptedException ex) {
throw new AnalysisException("Cannot segment pages!", ex);
}
this.computeDocumentOrientation(componentMap);
BxDocument output = new BxDocument();
BxPage[] pages = new BxPage[document.getPages().size()];
exec = Executors.newFixedThreadPool(PdfNLMContentExtractor.THREADS_NUMBER);
tasks = new ArrayList<Callable<NumBxPage>>();
int i = 0;
for (BxPage page : document.getPages()) {
tasks.add(new SingleSegmenter(page, i++));
}
try {
results = exec.invokeAll(tasks);
exec.shutdown();
for (Future<NumBxPage> result : results) {
NumBxPage p = result.get();
pages[p.index] = p.page;
}
for (BxPage p : pages) {
if (p.getBounds() != null) {
output.addPage(p);
}
}
return output;
} catch (ExecutionException ex) {
throw new AnalysisException("Cannot segment pages!", ex);
} catch (InterruptedException ex) {
throw new AnalysisException("Cannot segment pages!", ex);
}
}
private BxPage parsePageNode(Element elem) {
BxPage page = new BxPage();
double minX = 0, minY = 0, maxX = 0, maxY = 0;
boolean started = false;
List<Element> e = getChildren("Zone", elem);
for (Element zo : e) {
BxZone zon = parseZoneNode(zo);
page.addZone(zon);
BxBounds zoneBounds = zon.getBounds();
if (!started) {
minX = zoneBounds.getX();
minY = zoneBounds.getY();
maxX = zoneBounds.getX() + zoneBounds.getWidth();
maxY = zoneBounds.getY() + zoneBounds.getHeight();
started = true;
}
if (zoneBounds.getX() < minX) {
minX = zoneBounds.getX();
}
if (zoneBounds.getX() + zoneBounds.getWidth() > maxX) {
maxX = zoneBounds.getX() + zoneBounds.getWidth();
}
if (zoneBounds.getY() < minY) {
minY = zoneBounds.getY();
}
if (zoneBounds.getY() + zoneBounds.getHeight() > maxY) {
maxY = zoneBounds.getY() + zoneBounds.getHeight();
}
}
Collections.sort(
page.getZones(),
new Comparator() {
@Override
public int compare(Object t, Object t1) {
BxZone z1 = (BxZone) t;
BxZone z2 = (BxZone) t1;
int ret = Double.compare(z1.getBounds().getY(), z2.getBounds().getY());
if (ret == 0) {
ret = Double.compare(z1.getBounds().getX(), z2.getBounds().getX());
}
return ret;
}
});
return page.setBounds(new BxBounds(minX, minY, maxX - minX, maxY - minY));
}
private void appendPage(Document doc, Element parent, BxPage page, Object... hints)
throws TransformationException {
Element node = doc.createElement("Page");
appendPropertyIfNotNull(doc, node, "PageID", page.getId());
appendProperty(doc, node, "PageType", "");
appendProperty(doc, node, "PageNumber", "");
appendProperty(doc, node, "PageColumns", "");
appendPropertyIfNotNull(doc, node, "PageNext", page.getNextId());
appendProperty(doc, node, "PageZones", "");
for (BxZone zone : page.getZones()) {
appendZone(doc, node, zone, hints);
}
parent.appendChild(node);
}
@Override
public double calculateFeatureValue(BxLine line, BxPage page) {
if (!line.hasPrev()) {
return 1.0;
}
double avLength = 0;
int linesCount = 0;
for (BxZone zone : page.getZones()) {
for (BxLine l : zone.getLines()) {
linesCount++;
avLength += l.getBounds().getWidth();
}
}
if (linesCount == 0 || avLength == 0) {
return 0;
}
avLength /= linesCount;
return line.getPrev().getBounds().getWidth() / avLength;
}
@Override
public double calculateFeatureValue(BxZone zone, BxPage page) {
return page.getNext() == null ? 1.0 : 0.0;
}
@Override
public double calculateFeatureValue(BxLine object, BxPage context) {
return object.getBounds().getX() - context.getX();
}