Marmelator/ocr-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 64 Wiki Activity

RED-7669: improve ocr

Browse Source 
* fix pmd
This commit is contained in:
Kilian Schuettler 2024-05-13 11:35:57 +02:00
parent 18ba1daaef
commit 7b5a175440
2 changed files with 30 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
ocr-service-v1/ocr-service-processor/src/main/java/com/knecon/fforesight/service/ocr/processor/initializer/NativeLibrariesInitializer.java
View File

@ -34,12 +34,16 @@ public class NativeLibrariesInitializer {
System.setProperty("jna.library.path", System.getenv("VCPKG_DYNAMIC_LIB"));
log.info("Asserting Native Libraries loaded");
NativeLibrary leptonicaLib = NativeLibrary.getInstance("leptonica");
assert leptonicaLib != null;
log.info("Leptonica library loaded from {}", leptonicaLib.getFile().getAbsolutePath());
NativeLibrary tesseractLib = NativeLibrary.getInstance("tesseract");
assert tesseractLib != null;
log.info("Tesseract library loaded from {}", tesseractLib.getFile().getAbsolutePath());
try (NativeLibrary leptonicaLib = NativeLibrary.getInstance("leptonica")) {
assert leptonicaLib != null;
log.info("Leptonica library loaded from {}", leptonicaLib.getFile().getAbsolutePath());
}
try (NativeLibrary tesseractLib = NativeLibrary.getInstance("tesseract")) {
assert tesseractLib != null;
log.info("Tesseract library loaded from {}", tesseractLib.getFile().getAbsolutePath());
}
}
}

31
ocr-service-v1/ocr-service-processor/src/main/java/com/knecon/fforesight/service/ocr/processor/service/OcrResultWriter.java
View File

@ -85,15 +85,16 @@ public class OcrResultWriter {
}
@SuppressWarnings("PMD")
private List<Rectangle2D> getTextBBoxes(Page page) {
List<Rectangle2D> textBBoxes = new ArrayList<>();
try (var textExtractor = new TextExtractor()) {
textExtractor.begin(page);
try {
for (TextExtractor.Line line = textExtractor.getFirstLine(); line.isValid(); line = line.getNextLine()) {
for (var word = line.getFirstWord(); word.isValid(); word = word.getNextWord()) {
for (TextExtractor.Line line = textExtractor.getFirstLine(); line.isValid(); line = getNextLine(line)) {
for (TextExtractor.Word word = line.getFirstWord(); word.isValid(); word = getNextWord(word)) {
textBBoxes.add(Converter.toRectangle2D(word.getBBox()));
}
}
@ -105,9 +106,19 @@ public class OcrResultWriter {
}
private static Function<Integer, Integer> pageNumber1IdxTo0IdxMapper() {
// PDFBox uses a 0-based index for page numbers internally, while we use a 1-based index
return p -> p - 1;
private static TextExtractor.Word getNextWord(TextExtractor.Word word) {
TextExtractor.Word nextWord = word.getNextWord();
word.close();
return nextWord;
}
private static TextExtractor.Line getNextLine(TextExtractor.Line line) {
TextExtractor.Line newLine = line.getNextLine();
line.close();
return newLine;
}
@ -133,7 +144,6 @@ public class OcrResultWriter {
}
private VisualizationsOnPage createDebugTextVisualizations(List<OcrResultToWrite> ocrResultsToWrite, List<Rectangle2D> textBBoxes) {
List<TextPositionInImage> wordsToDraw = new ArrayList<>();
@ -175,7 +185,6 @@ public class OcrResultWriter {
}
private VisualizationsOnPage createDebugBBoxVisualizations(List<OcrResultToWrite> ocrResultsToWrite) {
List<TextPositionInImage> words = ocrResultsToWrite.stream()
@ -199,9 +208,9 @@ public class OcrResultWriter {
private List<ColoredLine> quadPointAsLines(QuadPoint rect) {
return List.of(new ColoredLine(new Line2D.Double(rect.a(), rect.b()), Color.ORANGE, 1),
new ColoredLine(new Line2D.Double(rect.b(), rect.c()), Color.BLUE, 1),
new ColoredLine(new Line2D.Double(rect.c(), rect.d()), Color.GREEN, 1),
new ColoredLine(new Line2D.Double(rect.d(), rect.a()), Color.MAGENTA, 1));
new ColoredLine(new Line2D.Double(rect.b(), rect.c()), Color.BLUE, 1),
new ColoredLine(new Line2D.Double(rect.c(), rect.d()), Color.GREEN, 1),
new ColoredLine(new Line2D.Double(rect.d(), rect.a()), Color.MAGENTA, 1));
}