Hi,
PDFs are not the best entity to be used as a data source due to its closed and complex internal structure.
I think that it may help if you let us know the actual goal of your test.
Possible options:
-- You are executing the test as a regression using stable static data in order to verify that same results are generated. Obviously in this case you may store images of the key areas of the report and do image search/comparison in the test.
-- If the test is executed against dynamic data, produces dynamic result and it is out of test scope to verify how raw source data is converted into PDF (e.g. some third-party component is used to generate the PDF), then you may talk to developers and ask them to provide you with the raw data used to generate the report and verify data correctness. Than, optionally, you may have additional step that will verify report layout without paying attention to figures. This layout verification can be done either manually or using some visual testing tool (e.g. Eyes by Applitools).
The most difficult approach is to use some pdf-parsing library (e.g. PDFBox) and parse the report to get data from it. The success of this approach significantly depends on how PDF file was generated.
Some PDFBox-related links:
http://community.smartbear.com/t5/Functional-Web-Testing/Getting-exception-while-trying-to-read-PDF-file-using-pdfbox-dll/td-p/85511
https://community.smartbear.com/t5/TestComplete-Desktop-Testing/Compare-pdf-files/m-p/122339
http://community.smartbear.com/t5/Functional-Web-Testing/how-to-get-the-content-of-a-pdf-file-on-window/td-p/60183
http://community.smartbear.com/t5/Functional-Web-Testing/Unable-to-read-pdf-dotNET-org-apache-pdfbox-pdmodel-PDDocument/td-p/94722