You Are Cleared for Access... 1.2 2004/12/18 01:11:48 US/Central 2004/12/19 16:43:01.731 US/Central Scott Harrison cronoman@rice.edu Jeremy Christopher Beasley prodig@rice.edu Brent Patrick Carroll bcarroll@rice.edu Richard G. Baraniuk richb@rice.edu Scott Harrison cronoman@rice.edu Jeremy Christopher Beasley prodig@rice.edu Brent Patrick Carroll bcarroll@rice.edu Richard G. Baraniuk richb@rice.edu Fingerprint Image Matching This module refers to all of our results, problems we had, and conclusions we have made. Also included are acknowledgements.

Scanning... From all the methods we used to identify and match images, we obtained the following results:

Results of Various Matching Processes These are the correlations between various reference images (A1-D1) and our scanned images.

that the Minutiae-Based Approach was not considered in the results because of time issues and complexity. All of the results seem pretty similar except for the calculation times and the "False Positive" correlation. That is, depending on the image and method, the number of images that reported a "match" in the case of a non-match decreased between the two equivalent domain methods, and the Rotation+FFT Method. Image D1 also seems to be a fluke, since it reported perfect correlation and false positve matching.

Error. You are not in the database. Fingerprint matching is not always a straightforward process with definite steps. Many different problems can arise to hamper the process such as, lack of proper scanning equipment, slow computers, and sheer size of a database. The main problem that arose, however, was the implementation of the Minutiae-based Method. This method is more along the lines of what security agencies like the FBI use, and their technology and algorithms are much more sophisticated than the simple tools we have to use. The complexity of implementing such a process was beyond our programming capabilities in Matlab and beyond our time frame given to work. Given more time and resources, our group might have been able to implement such a method.

Match Found. You May Proceed... As a result of our research we have come to the following conclusions: Image quality plays a huge role in match probability Frequency domain analysis is generally faster for image recognition processes The addition of the rotational matrix improved match probability and decreased error since it sweeps over various orientations of image Some suggestions for further research include better classification algorithms, higher resolution scanners, and better unique feature data extraction. The most "optimal" method of matching images would be a combination of the Minutiae-based approach and the Rotation+FFT Method. This, however, can only be had with better technology and time. The normal tradeoff in the matching process is Time to Complete vs. Efficiency/Precision in Matching. When we consider this fact, the optimal case to use in signal processing matching is the Rotation + FFT Method .

Acknowledgements We would like to thank and acknowledge the following people and sources: Rice Univ. Dept. of Electrical Engineering: Dr. Richard Baraniuk, William Chan, Chris Rozell Previous Elec 301 Groups (2002 - Fresh Prints of Rice, 2000 - Fingerprint Matching) Cain Project: Dr. Linda Driskil National Institute of Standards and Technology Pattern Recognition and Image Processing Lab Department of Computer Science And Engineering Michigan State University: Umut Uludag and Anil Jain Group Fingerprint: Scott Harrison, Brent Carroll, and Jeremy Beasley