Page 261 - Automated Fingerprint Identification Systems (AFIS)
P. 261

246  AUTOMATED FINGERPRINT IDENTIFICATION SYSTEMS



                              new latent cog record entered into the database, it would be possible for a poor-
                              quality image with only a few minutiae points to spawn hundreds of false can-
                              didates, particularly if it was saved as a “cold” case. At times, the sheer volume
                              of tenprint to unsolved candidates awaiting verification could be daunting.
                              Examiners had to ask whether the time spent in reviewing hundreds of TP/UL
                              candidates produced more identifications than reviewing only a few TP/UL
                              candidates by making the search more restrictive and using the remaining time
                              for latent to tenprint (LT/TP) searches. This was an individual as well as a man-
                              agement decision.



                              10.4 SYSTEM-WIDE UPGRADE

                              In 1998, DCJS embarked on another bold plan to improve SAFIS through a
                              system-wide upgrade. The multi-year program, under the direction of Deputy
                              Commissioner Dan Foro and Chief of Biometric Identification Jack Meagher,
                              had several enormous tasks, such as the following:

                              1. Recode the entire tenprint database.
                              2. Recode the latent cog database.
                              3. Install and test new operating and application software.
                              3. Install new coders.
                              4. Install new matchers.
                              5. Install new workstation terminals.

                                 In the 10 years since the original conversion of the DCJS database, digital
                              coding algorithms had greatly improved, as had the matching algorithms. The
                              operating system software was faster and more reliable. Critical operating
                              software and hardware components were available as commercial-off-the-shelf
                              (COTS) items. There was less equipment required for faster processing.
                                 To accomplish these objectives, while the day-to-day work at DCJS continued,
                              copies of the tenprint and latent cog image databases were sent to Sagem for
                              recoding. Samples of the recoded images were checked for accuracy and reli-
                              ability. The improved coders found minutiae where the older coders did not,
                              and more reliably distinguished between clear and marginal minutiae. A new
                              random array of independent drives (RAID) storage system was introduced,
                              which provided images faster and more reliably than in the past. New tenprint
                              and latent print workstations that had the newer coders were installed. As of
                              July 1999, all new tenprint images were coded using the new coders. Likewise,
                              latent print examiners also had the benefit of new coders on their workstations
                              to more clearly identify minutiae and other image characteristics. By the end
                              of 1999, the new SAFIS was fully in place. Accuracy tests of both the tenprint
   256   257   258   259   260   261   262   263   264   265   266