While it may seem easy enough to look at a set of fingerprints and tell if it has been altered, Jain says it’s not practical.
Graduate student Soweon Yoon helped design a precise algorithm which helps software systems identify altered prints.
“Some of these databases are extremely, extremely large. So the FBI database is an example, has about 70 million subjects. That means there are 700 million fingerprint images if each person has 10 fingers. And so there is no way anybody could manually check whether there are altered fingerprints or not.”
Eliminating the guesswork
As it analyzes images, the application does not identify people, it simply alerts officials to an anomaly, so they can perform a secondary inspection. And, Jain says, it also takes the guesswork out of determining whether or not a print has been altered.
“Different human experts will have different degrees of proficiency in identifying (prints), and they get tired and so their decisions are not necessarily going to be consistent. Whereas a computer algorithm is going to be consistently behaving, there is no subjectivity in it once you tune it, it’s going to behave in the same way.”
The application also helps distinguish between accidental and intentional alterations. Most accidents result in scars on only a few fingers. If there are changes in five or more, that was most likely done on purpose.
最新
2013-11-25
2013-11-25
2013-11-25
2013-11-25
2013-11-25
2013-11-25