从 PDFBox 中剥离时的文本坐标

Question

我正在尝试使用 PDFBox 从 pdf 文件中提取带有坐标的文本。

我混合了一些在互联网上找到的方法/信息（stackoverflow 也是），但我的坐标问题似乎不正确。例如，当我尝试使用坐标在 tex 上绘制矩形时，矩形会在其他地方绘制。

这是我的代码（请不要评判风格，写得很快就是为了测试）

TextLine.java

    import java.util.List;
    import org.apache.pdfbox.text.TextPosition;

    /**
     *
     * @author samue
     */
    public class TextLine {
        public List<TextPosition> textPositions = null;
        public String text = "";
    }

myStripper.java

    import java.io.IOException;
    import java.util.ArrayList;
    import java.util.List;
    import org.apache.pdfbox.pdmodel.PDDocument;
    import org.apache.pdfbox.pdmodel.PDPage;
    import org.apache.pdfbox.text.PDFTextStripper;
    import org.apache.pdfbox.text.TextPosition;

    /*
     * To change this license header, choose License Headers in Project Properties.
     * To change this template file, choose Tools | Templates
     * and open the template in the editor.
     */

    /**
     *
     * @author samue
     */
    public class myStripper extends PDFTextStripper {
        public myStripper() throws IOException
        {
        }

        @Override
        protected void startPage(PDPage page) throws IOException
        {
            startOfLine = true;
            super.startPage(page);
        }

        @Override
        protected void writeLineSeparator() throws IOException
        {
            startOfLine = true;
            super.writeLineSeparator();
        }

        @Override
        public String getText(PDDocument doc) throws IOException
        {
            lines = new ArrayList<TextLine>();
            return super.getText(doc);
        }

        @Override
        protected void writeWordSeparator() throws IOException
        {
            TextLine tmpline = null;

            tmpline = lines.get(lines.size() - 1);
            tmpline.text += getWordSeparator();

            super.writeWordSeparator();
        }


        @Override
        protected void writeString(String text, List<TextPosition> textPositions) throws IOException
        {
            TextLine tmpline = null;

            if (startOfLine) {
                tmpline = new TextLine();
                tmpline.text = text;
                tmpline.textPositions = textPositions;
                lines.add(tmpline);
            } else {
                tmpline = lines.get(lines.size() - 1);
                tmpline.text += text;
                tmpline.textPositions.addAll(textPositions);
            }

            if (startOfLine)
            {
                startOfLine = false;
            }
            super.writeString(text, textPositions);
        }

        boolean startOfLine = true;
        public ArrayList<TextLine> lines = null;

    }

AWT 按钮上的点击事件

 private void jButton1MouseClicked(java.awt.event.MouseEvent evt) {                                      
    // TODO add your handling code here:
    try {
        File file = new File("C:\\Users\\samue\\Desktop\\mwb_I_201711.pdf");
        PDDocument doc = PDDocument.load(file);

        myStripper stripper = new myStripper();

        stripper.setStartPage(1); // fix it to first page just to test it
        stripper.setEndPage(1);
        stripper.getText(doc);

        TextLine line = stripper.lines.get(1); // the line i want to paint on

        float minx = -1;
        float maxx = -1;

        for (TextPosition pos: line.textPositions)
        {
            if (pos == null)
                continue;

            if (minx == -1 || pos.getTextMatrix().getTranslateX() < minx) {
                minx = pos.getTextMatrix().getTranslateX();
            }
            if (maxx == -1 || pos.getTextMatrix().getTranslateX() > maxx) {
                maxx = pos.getTextMatrix().getTranslateX();
            }
        }

        TextPosition firstPosition = line.textPositions.get(0);
        TextPosition lastPosition = line.textPositions.get(line.textPositions.size() - 1);

        float x = minx;
        float y = firstPosition.getTextMatrix().getTranslateY();
        float w = (maxx - minx) + lastPosition.getWidth();
        float h = lastPosition.getHeightDir();

        PDPageContentStream contentStream = new PDPageContentStream(doc, doc.getPage(0), PDPageContentStream.AppendMode.APPEND, false);

        contentStream.setNonStrokingColor(Color.RED);
        contentStream.addRect(x, y, w, h);
        contentStream.fill();
        contentStream.close();

        File fileout = new File("C:\\Users\\samue\\Desktop\\pdfbox.pdf");
        doc.save(fileout);
        doc.close();
    } catch (Exception ex) {

    }
}                                     

有什么建议吗？我做错了什么？

Answer 1

以下代码对我有用：

    // Definition of font baseline, ascent, descent: https://en.wikipedia.org/wiki/Ascender_(typography)
    //
    // The origin of the text coordinate system is the top-left corner where Y increases downward.
    // TextPosition.getX(), getY() return the baseline.
    TextPosition firstLetter = textPositions.get(0);
    TextPosition lastLetter = textPositions.get(textPositions.size() - 1);

    // Looking at LegacyPDFStreamEngine.showGlyph(), ascender and descender heights are calculated like
    // CapHeight: https://stackoverflow.com/a/42021225/14731
    float ascent = firstLetter.getFont().getFontDescriptor().getAscent() / 1000 * lastLetter.getFontSize();
    Point topLeft = new Point(firstLetter.getX(), firstLetter.getY() - ascent);

    float descent = lastLetter.getFont().getFontDescriptor().getDescent() / 1000 * lastLetter.getFontSize();
    // Descent is negative, so we need to negate it to move downward.
    Point bottomRight = new Point(lastLetter.getX() + lastLetter.getWidth(),
        lastLetter.getY() - descent);

    float descender = lastLetter.getFont().getFontDescriptor().getDescent() / 1000 * lastLetter.getFontSize();
    // Descender height is negative, so we need to negate it to move downward
    Point bottomRight = new Point(lastLetter.getX() + lastLetter.getWidth(),
        lastLetter.getY() - descender);

换句话说，我们正在创建一个从字体的上升到下降的边界框。

如果您想以左下角的原点呈现这些坐标，请参阅https://stackoverflow.com/a/28114320/14731 了解更多详细信息。您需要应用这样的转换：

contents.transform(new Matrix(1, 0, 0, -1, 0, page.getHeight()));

Answer 2

这只是过度PdfTextStripper 坐标归一化的另一种情况。就像您一样，我曾认为通过使用TextPosition.getTextMatrix()（而不是getX() 和getY）可以获得实际坐标，但是不，即使这些矩阵值也必须更正（至少在PDFBox 2.0.x 中，我没有检查 1.8.x)，因为矩阵乘以平移，使裁剪框的左下角成为原点。

因此，在您的情况下（裁剪框的左下角不是原点），您必须更正这些值，例如通过替换

        float x = minx;
        float y = firstPosition.getTextMatrix().getTranslateY();

通过

        PDRectangle cropBox = doc.getPage(0).getCropBox();

        float x = minx + cropBox.getLowerLeftX();
        float y = firstPosition.getTextMatrix().getTranslateY() + cropBox.getLowerLeftY();

而不是

你现在得到

但显然，您还必须稍微修正高度。这是由于PdfTextStripper 确定文本高度的方式：

    // 1/2 the bbox is used as the height todo: why?
    float glyphHeight = bbox.getHeight() / 2;

（来自LegacyPDFStreamEngine中的showGlyph(...)，PdfTextStripper的父类）

虽然字体边界框确实通常太大，但通常一半是不够的。