Instructions

U.are.U SDK - Developer Guide 16
Developing Applications
Defining the Data Retention Policy
After a fingerprint scan, an FID is created, which contains the actual image. Each fingerprint image takes roughly
140K of storage. With modern computers, that is not a huge amount, but each enrolled user may have several
fingerprints scanned. Multiplied by the number of potential users, this can add up to a fair amount of data.
To identify fingerprints, you must extract the fingerprint characteristics to create an FMD, which is < 1.5K per
fingerprint.
Some applications choose to retain the full image records, but other companies discard the image record and retain
only the minutiae data in the form of an FMD.
IMPORTANT: If you discard the image record, you cannot reconstruct the original fingerprint image from the FMD,
the FMD is only useful for identifying/verifying fingerprints.
From time to time Crossmatch may release a new version of the U.are.U SDK that will provide improved accuracy in
the feature extraction process. If you do not retain the fingerprint images, you will not be able to redo the feature
extraction using the new version of the SDK/runtime in order to take advantage of these improvements.
Specifying the Fingers to Be Scanned
When you enroll people into your system, you will usually want more than one finger enrolled. This allows for injury
and makes it easier for people who have fingerprints that are difficult to recognize. For some applications you may
want to scan all ten fingers or take multiple fingerprint impressions for individual fingers.
A typical policy would be to require both index fingers or both thumbs to be scanned. Thumbs are typically the worst
in terms of recognition accuracy and thumbs require different capture devices with larger area and different
ergonomics. Where possible, avoid any industrial design forcing users to use thumbs.
The preferred approach is to enroll, at a minimum, both index and both middle fingers. Middle fingers are usually the
best, probably because people have almost as good dexterity with middle fingers as with index fingers, and yet middle
fingers have fingerprints that are typically less worn than index fingers.
Some solutions are ergonomically designed in a way that only the right or only the left hand can be used conveniently.
In this case the right hand is better (probably because the majority of people are right handed), and at least three fingers
need to be enrolled: index, middle and ring fingers.
Ergonomics and the correct finger placement are extremely important. Poor ergonomics can easily increase the false
negative identification rate by a factor or 5 to 10. The system users need to be aware of correct finger placement for
best results.
Optimizing Fingerprint Applications
To identify a fingerprint against a large database can take a considerable amount of time and create unacceptable
delays between the fingerprint scan and the user authorization. Identifying fingerprints will be faster if the database of
fingerprints is in memory rather than retrieved from disk. If you have a large database of users, you may need to
provision your server with an appropriate amount of RAM to handle the searches.
This SDK is not optimized for large scale identification. If you are developing such an application, you may want to
contact Crossmatch to get help selecting the technology that will best fit your needs.
To ensure the fastest response time, you must weigh whether it will be faster to transmit the image record to a server
for FMD extraction or whether it is faster to extract the FMD on the reader and transmit only the FMD to the server for
identification/verification.
For readers that are used by a limited number of people (e.g., kiosks or pharmacy cabinets), you may have the device
identify fingerprints against a limited set of FMDs. However this requires that you keep the FMDs in the device in sync
with your central database, to ensure that new employees are able to gain access and departing employees’ privileges
are revoked quickly.