Detecting if a PDF file contains 3D element

Question

Is there any way to detect if a PDF contains 3D element (Universal 3D object embedded) without reading the file contents? Can this information be obtained from metadata?

Answer 1

AFAIK, there is no requirement to put any info into its metadata about the fact that 3D elements may be contained in the document.

Some U3D writing software may put some hint about it into the XML metadata though.

Long Answer

You'll have to parse the PDF page tree at least partially in order to find out.

Technically, 3D elements are implemented as annotations. To discover any annotation, you'd have to follow this parsing path:

1. Read the trailer. It tells you the object number of the /Root indirect object of the document.

2. Read the cross reference table. It tells you the byte offsets for each indirect object within the document.

3. Go to the /Root indirect object. Read its /Pages key. This tells you which indirect object represents the root of the document's page tree.

4. Go to the indirect object which represents the /Pages. Read its /Kids key. This tells you which other indirect objects represent document pages.

5. Go to each indirect object representing a document page. Look for any (optionally present) /Annots key. If present it will point to other indirect objects representing (possibly all sorts of) annotations.

Now you've found out if the PDF contains annotation(s) or not. If not, stop here. If yes, go on to determine the annotation type(s):

1. Go to all indirect objects found in the last step. They are of /Type /Annot. See if they are additionally of /Subtype /3D. If yes, you have found a 3D annotation. (Attention, this may still not be a U3D one!)

2. Within the lastly found indirect object(s) -- the one(s) with the /Subtype /3D key -- look for an additional key of /3DD. It points to this indirect object which contains the actual 3D stream.

3. Go to the indirect object containing the 3D stream. Its object dictionary should again contain a key:value pair of /Type /3D. Look at its /Subtype key. If it says /U3D you have found what you looked for...

Short Answer

You may be lucky and harvest some low-hanging fruits by using good old grep like this:

$> grep -a U3D cc-7-july09.pdf

  /Subtype /U3D
  /MS /U3D
  /U3DPath [ <135BB3D42FBD85F7C2E178> <056D9A891FB5FDCE8E> ]
  /MS /U3D
  /U3DPath [ <5FFAF35CE3CBD34FAE5360> <4DDFD6048FC6DA05> ]
  /MS /U3D
  /U3DPath [ <2E4E4FD7FEC771038BC5EA> <2A6579CC91BE0B> ]
  /MS /U3D
  /U3DPath [ <6F303AF9850721D5D1FC6C> <7D1B08BEAE4A5A9BEDBB> ]
  /MS /U3D
  /U3DPath [ <F270A04603F0DE08B8AA29> <EE5180016FFBD542> ]
  /MS /U3D
  /U3DPath [ <A1D5848F6841ADA9A3583C> <A3F8A5D45849D392EF> ]
  /MS /U3D
  /U3DPath [ <34B8650D178BBDFF61DC03> <2D8F4C7D3CD980F976> ]
  /MS /U3D
  /U3DPath [ <843CD0339FD1852CCA235B> <9719FB65A990897F> ]

However, this will not work for all 3D PDF documents, especially if the 3D elements are part of an object stream.

Answer 2

For cases where it is sufficient to know if a PDF contains any 3D content (including, but not being exclusive to Universal 3D objects), you could also use the VeraPDF software in feature extraction mode. Follow the steps below to get a list of all Annotation types (including 3D annotations).

First edit VerapDF's "features.xml" configuration file as explained here:

https://docs.verapdf.org/cli/config/#features.xml

Make sure that <feature>ANNOTATION</feature> is included in the enabledFeatures element.

Using this file as an example, run:

verapdf --off --extract action_goto3dview.ar10.pdf > action_goto3dview.ar10.xml

In the output, check the "annotations" element, which lists all annotations that are present in the file, and look for annotations with subType "3D" (which indicates a 3D annotation):

<annotation id="annotIndir186">
<subType>3D</subType>
<rectangle lly="129.348" llx="163.939" urx="437.813" ury="331.861"></rectangle>
<width>273.874</width>
<height>202.513</height>
<contents>3D Model</contents>
<annotationName>3D3</annotationName>
<resources>
    <xobject id="xobjIndir187"></xobject>
</resources>
<invisible>false</invisible>
<hidden>false</hidden>
<print>true</print>
<noZoom>false</noZoom>
<noRotate>false</noRotate>
<noView>false</noView>
<readOnly>true</readOnly>
<locked>false</locked>
<toggleNoView>false</toggleNoView>
<lockedContents>false</lockedContents>
</annotation>

This corresponds to Step 1 in @kurt-pfeifle's answer above. As VeraPDF doesn't drill down to the level required to identify U3D streams, I'd suggest @kurt-pfeifle's answer for cases where that level of detail is needed.

Answer 3

For anyone having the same problem as us, this is an approach we come up with using "iText" (free version is still avaiable).

The downside is that you will need to iterate every page through the file to check the content but it's still fast enough for us.

        PdfReader reader = new PdfReader(contents);
        int pages = reader.getNumberOfPages();
        boolean pdf3D = false;
        for (int i = 1; i <= pages; i++) {
            PdfDictionary page = reader.getPageN(i);
            PdfArray array = page.getAsArray(PdfName.ANNOTS);
            if (array == null) {
                continue;
            }
            for (ListIterator<PdfObject> iter = array.listIterator(); iter.hasNext();) {
                PdfDictionary annot = (PdfDictionary) PdfReader.getPdfObject(iter.next());
                PdfObject pdfObject = annot.get(PdfName.SUBTYPE);
                if (pdfObject != null) {
                    if (PdfName._3D.equals(pdfObject) || PdfName.GOTO3DVIEW.equals(pdfObject)) {
                        pdf3D = true;
                        break;
                    }
                }
            }
            if (pdf3D) {
                // if we already any of 3D element, we can break the loop
                break;
            }
        }