EE 7740 Fingerprint Recognition Bahadir K Gunturk Biometrics
EE 7740 Fingerprint Recognition Bahadir K. Gunturk
Biometrics n n Biometric recognition refers to the use of distinctive characteristics (biometric identifiers) for automatically recognition individuals. These characteristics may be q q n Physiological (e. g. , fingerprints, face, retina, iris) Behavioral (e. g. , gait, signature, keystroke) Biometric identifiers are actually a combination of physiological and behavioral characteristics, and they should not be exclusively identified into either class. (For example, speech is determined partly by the physiology and partly by the way a person speaks. ) Bahadir K. Gunturk 2
Biometrics Bahadir K. Gunturk 3
Biometrics Bahadir K. Gunturk 4
Biometrics Bahadir K. Gunturk 5
Fingerprint n Human fingerprints have been discovered on a large number of archeological artifacts and historical items. Bahadir K. Gunturk 6
Fingerprint n In 1684, an English plant morphologist published the first scientific paper reporting his systematic study on the ridge and pore structure in fingerprints. Bahadir K. Gunturk 7
Fingerprint Bahadir K. Gunturk 8
Fingerprint n A fingerprint image may be classified as q Offline: n q Inked impression of the fingertip on a paper is scanned Live-scan: n Optical sensor, capacitive sensors, ultrasound sensors, … Critical parameter are: Resolution, area, contrast, noise, geometric accuracy. Bahadir K. Gunturk 9
Fingerprint n n The fingerprint pattern exhibits different types of features. At the global level, the ridge line flow has one the following patterns. Singular points are sort of control points around which a ridge line is “wrapped”. There are two types of singular points: loop and delta. However, these singular points are not sufficient for accurate matching. Bahadir K. Gunturk 10
Fingerprint n n At the local level, there different local ridge characteristics. The two most prominent ridge characteristics, called minutiae, are: q q n Ridge termination Ridge bifurcation At the very-fine level, intra-ridge details (sweat pores) can be detected. They are very distinctive; however, very high-resolution images are required. Termination Bifurcation Bahadir K. Gunturk 11
Example n Matching is not easy due to: displacement, rotation, partial overlap, nonlinear distortion, changing skin condition, noise, feature extraction errors, etc. Bahadir K. Gunturk 12
Example n There are many “ambiguous” fingerprints, whose exclusive membership cannot be reliably stated even by human experts. Bahadir K. Gunturk 13
Fingerprint Recognition Approaches n n n Correlation-based matching: Intensity based correlation between the fingerprint images are computed. Minutiae-based matching: Minutiae are extracted from two fingerprints and stored as sets of points in the 2 D plane. Matching is done based on minutiae pairings. Ridge feature-based matching: Local orientation and frequency of ridges, ridge shape, texture, etc are used for matching. Bahadir K. Gunturk 14
Minutiae Detection n Binarize the image (using global thresholding, local thresholding, etc. ) Apply thinning (by, for example, using morphological operations) to get the skeleton image. Analyze the neighborhood of each pixel in the skeleton image. Bahadir K. Gunturk 15
Minutiae Detection n Minutia detection may be followed by post-processing to remove false minutiae structures. Bahadir K. Gunturk 16
Fingerprint Matching Bahadir K. Gunturk 17
Fingerprint Matching Bahadir K. Gunturk 18
Fingerprint Matching Bahadir K. Gunturk 19
Performance • Comparison • Fingerprints [FVC 2002] • False reject rate: 0. 2% • False accept rate: 0. 2% • Face [FRVT 2002] • False reject rate: 10% • False accept rate: 1% • Voice [NIST 2000] • False reject rate: 10 -20% • False accept rate: 2 -5% Bahadir K. Gunturk 20
Performance • How to improve • Fingerprint enhancement • Estimating deformations • Multiple matchers & combine results • Multimodel biometrics Bahadir K. Gunturk 21
- Slides: 21