I want to use PDFBox to get the texts from the pdf with their bounding boxes. I’ve managed to get together a code from here that does almost this, but as you can see the bounding boxes I get (drawn with blue) are not the correct size as if I were to select the texts. Where could the calculation go wrong in my code?

class CustomPDFTextStripper : PDFTextStripper() {

@Throws(IOException::class)
override fun writeString(text: String, textPositions: List<TextPosition>) {
    // Initialize bounding box coordinates
    val wordSeparator = wordSeparator
    val word: MutableList<TextPosition> = ArrayList()

    // Get the page height to correctly adjust the Y-coordinates
    val page = document.getPage(0)
    val pageHeight = page.mediaBox.height

    for (text in textPositions) {
        val thisChar = text.unicode
        if (thisChar != null && thisChar.isNotEmpty()) {
            if (thisChar != wordSeparator) {
                word.add(text)
            } else if (word.isNotEmpty()) {
                printWord(word, pageHeight)
                word.clear()
            }
        }
    }
    if (word.isNotEmpty()) {
        printWord(word, pageHeight)
    }
}

@Throws(IOException::class)
fun printWord(word: List<TextPosition>, pageHeight: Float) {
    if (word.isEmpty()) return

    // Create a bounding box for the word
    var boundingBox: Rectangle2D? = null
    for (text in word) {
        val box = Rectangle2D.Float(text.xDirAdj, pageHeight - text.yDirAdj - text.heightDir, text.widthDirAdj, text.heightDir)
        if (boundingBox == null) {
            boundingBox = box
        } else {
            boundingBox.add(box)
        }
    }

    // Draw the bounding box
    val page = document.getPage(0) // Assumes drawing on the first page
    PDPageContentStream(document, page, PDPageContentStream.AppendMode.APPEND, true, true).use { contentStream ->
        contentStream.setStrokingColor(Color.BLUE)
        contentStream.setLineWidth(1f)
        if (boundingBox != null) {
            contentStream.addRect(boundingBox.x.toFloat(), boundingBox.y.toFloat()+boundingBox.height.toFloat(), boundingBox.width.toFloat(), boundingBox.height.toFloat())
            contentStream.stroke()
        }
    }

    // Print word and bounding box details
    val builder = StringBuilder()
    for (text in word) {
        builder.append(text.unicode)
    }
    println("${builder.toString()} [(X=${boundingBox!!.x}, Y=${boundingBox.y}) height=${boundingBox.height} width=${boundingBox.width}]")
}

}

The result: