In order to have a first introduction to fingerprints, I watch a presentation made by Dr. Anil K. Jain. He is a University Distinguished Professor at Michigan State University, and did a lot of research on biometrics and especially fingerprints. The video and the slides are available here.
This video help me to have an overview of the work and research I need to do. Here are some points I found important. This slide shows the classic process for identification:
- in the first step (enrollment), the digit is scanned, and minutiae points are extracted. Then, only this information is stored in a database, not the whole scanned image.
- in the second step (authentication), the digit is scanned and minutiae points are extracted again. Then those points are matched (or not), with an other set of points stored in the database.
When we extract minutiae points, we try to find as much points of singularity as we can. They are of called either ridge bifurcation or ridge ending. Actually, we store the position information (x,y) and the angle of the ridge. We may extract core and delta points, which are characteristic points where the flow field of the ridges changes quickly. We also may store the type of the fingerprint to improve the speed of the algorithm.
Why do we use minutiae point matching and not image matching?
Most of the information in image is contained in the minutiae points. Every scan of a digit are different. It is really hard to match two images since there can be linear transformation between the two scans (rotation, translation, scaling and shear), plus other non linear transformation (elasticity of the skin, digit pressure), and noise coming from the sensor and the dryness of the finger.
I am currently learning how the RANSAC algorithm works, why it is used, and the mathematical background of it. I read the paragraph explaining the process in the book « Computer Vision: a modern approach », from David A. Forsyth and Jean Ponce (page 346). I need to go deeper in the details, so I am going to read the Handbook of Fingerprint Recognition from D. Maltoni, D. Maio, A. K. Jain, S. Prabhakar, and other web references.
3 comments:
Hello Sir,
Can u give me this code ransac for fingerprints identification. Right now i am working on Fingerprint verification using sampling algorithms,
Thanks in advance
The slides are removed.
Could you send me the slides?
Hello sir...Me a student n working on Fingerprint identification for my project.i have gone through minutiae detecting n storing em as a template. And for false removal and matching i need to give an id for all de thinned ridges which am stuck up n coudnt move forward...so it ll be a great help if u gimme an idea r an algorithm r respective code for doing dis operation...Thanx in Advance
Post a Comment