Process and OCR Results

Mismatch Confidence Thresholds

A minor difference is not always considered a mismatch. In certain circumstances, the difference is allowed. For example, if the barcode data contains the word North, while the OCR text contains the letter N, then the mismatch is an explainable discrepancy. If the barcode data contains a zero, while the OCR text interprets the zero as a letter O, then the mismatch is also an explainable discrepancy. Or a discrepancy might occur due to poor image quality as a result of factors such as blurriness or glare.

These types of discrepancies are accounted for with an OCR mismatch confidence threshold that is set by Intellicheck. For example, a threshold of 70 means there must be 70 percent or greater certainty that the two values are equivalent. In such a scenario, the confidence threshold operates as follows:

• If the certainty falls below 70, then the two values are color-coded as a mismatch.

• If the certainty falls at or above 70, then the two values are not color-coded a mismatch.

In summary, a mismatch will not automatically cause a color-code change, an OCR Mismatch status, or a Failed status. Mismatch confidence thresholds play a role in the analysis. Intellicheck sets the confidence thresholds; therefore, they are not configurable. Keep this in mind as you review results.

Results Example - North American Driver's License/Identification Cards

The following illustration shows an example results page for a North American driver license.

Several results are circled and numbered to illustrate scenarios where a mismatch or a discrepancy might be found.

1. Name. The barcode data did not match the OCR text. This is a mismatch. The barcode contains the full middle name, Jason. But the front image contains the first letter of the middle name, J. This mismatch might be considered an explainable difference.

2. Date of Birth. The barcode data matched the OCR text. This is not a mismatch. The difference is due to formatting, not the date itself.

3. Issue Date. The barcode did not match the OCR text. This is a mismatch. The barcode data contains the year 2017, while the OCR text contains the year 2016. Date discrepancies are always considered a mismatch. The confidence threshold for dates is 100 percent.

4. Is RealID. The barcode data did not match the OCR text. This is not a mismatch. The barcode does not contain RealID data. Issuing jurisdictions have different policies regarding RealID information.

5. Driver Class. The barcode data did not match the OCR text. This is a mismatch. The barcode contains the letter A, while the OCR text contains the number 4. This mismatch might be considered an OCR-read error.

6. Weight. The barcode data did not match the OCR text. This is not a mismatch. The barcode does not contain weight data, and neither does the OCR text. Missing data does not always indicate a mismatch.

Results Example - Passport Booklet

The following illustration shows an example results page for a United States passport.

Several results are circled and numbered to illustrate scenarios where a mismatch or a discrepancy might be found.

1. Name. This is a mismatch due to incomplete OCR information. The MRZ data did not match the OCR text. The MRZ contains the full name, HAPPY TRAVELER. But the OCR text contains the last name, TRAVELER.

2. Country Code. This is a mismatch due to missing OCR information. The MRZ data contained a country code of USA. But the OCR text did not define a country code.

3. Document Name. This is not a mismatch because there is no equivalent OCR text field for comparison.

Results Example - Other Documents (Passport Cards, Non-North American Driver's License/ID)

The following illustration shows an example results page for an international passport card.

Several results are circled and numbered to illustrate scenarios where a mismatch or a discrepancy might be found.

1. Name. This is a mismatch due to incomplete OCR information. The MRZ data did not match the OCR text. The MRZ contains the full name, HAPPY TRAVELER. But the OCR text contains the first name, HAPPY.

2. Expiration Date. This is not a mismatch, however, the date is expired. The Expired status icon is active (not shown).

3. Country Code. This is a mismatch due to missing OCR information. The MRZ data contained a country code of USA. But the OCR text did not define a country code.

4. Issuer. This is a mismatch due to missing OCR information. The MRZ data contained an issuer of USA. But the OCR text did not define an issuer.

5. Document Name. This is not a mismatch because there is no equivalent OCR text field for comparison.