Some of the things you might consider for pattern matching include when to use for page identification and good anchor patterns.
If your application must handle various page types that are similar, standard fingerprint identification (FindFingerprint) might generate mismatches. Pattern matching uses a smaller defined region of the page. You can select an area that is unique to each page type and thus avoid mismatches.
For pattern matching to work, you must have a good anchor pattern. If you are using geometric pattern matching, the image must be composed of simple solid regions. Avoid anything with shaded areas. The images in the following table provide some good and bad examples.
Good | Bad |
For text-based pattern matching, the only requirement is that the text pattern is unique to the page type.
The standard Fingerprint Matching action, FindFingerprint, can tolerate some minor page misalignment relative to the fingerprint image. For more information, see Auto registration with the FindFingerprint action. Pages that have more serious alignment problems or have inconsistently proportioned areas require pattern matching for registration. Faxed images are especially vulnerable because the sending and receiving fax machines might pull the paper through at different speeds, resulting in longer, or shorter images. Additionally, pages with OMR fields require accurate registration, especially if the boxes are closely spaced.
In certain situations, you might want to do image registration manually. You can register images by using the AIndex web client. For details, see Manual page identification and registration.