There have been many major advances in biometric technologies over the last few decades that have increased the security of our nation. Automated biometric systems have aided in the capture of some of the world’s most dangerous people. However, there is a vulnerability that potentially results in criminals, terrorists, and other persons of interest going undetected by even the most advanced biometric systems.
A growing number of biometric systems are making changes in order to expedite the biometric capture and search time. One such change removed the requirement of capturing rolled fingerprints in lieu of only capturing the flat fingerprints in just three images: left four fingers, right four fingers and two thumbs. These transactions became known in the biometric community as the 4-4-2 captures or simply IDFlat.
The benefits of the IDFlat transactions do not come without a cost. When rolled prints were present, if the segmentation algorithm failed, the sequence check would also fail, resulting in a rejection or human review. Without rolled prints, inaccurate segmentation goes undetected, leading to missed matches.
Many segmentation algorithms in production today were evaluated by NIST under the Slap Fingerprint Segmentation Evaluation (SlapSeg) or SlapSeg II. These evaluations only evaluate algorithms using images that are captured at the correct orientation, not evaluating their ability to handle extreme angles.
Below is an example of the NIST Fingerprint Segmentation (NFSEG) algorithm failing to segment a right-four finger slap that was captured at an angle.
Lakota has developed a fingerprint segmentation algorithm to address these issues. Our state-of-the-art segmentation algorithm is capable of detecting and segmenting fingerprints from images regardless of the angle captured. Below is a right four-finger slap image segmented using Lakota’s new segmentation algorithm. Note, this is the same image that failed to segment using NFSEG above.
The image below demonstrates the ability of the Lakota segmentation algorithm to correctly segment fingerprint images even when the slap was captured upside-down. This algorithm can accurately detect the orientation of the fingerprints at a full 360-degrees of rotation.