A multi-level fingerprint feature extraction and verification system using pores

The use of biometrics, particularly fingerprints, as a means of identity verification has become an important field of research. Current fingerprint verification systems only take advantage of a small amount of the unique information present in fingerprint images. The consideration of additional discriminatory information, particularly content from the third level of fingerprint detail, has been largely avoided due to claims that sensors are unable to reliably detect the information. While this was true in the past, modern sensors are capable of detecting some of this information, particularly the location of pores, with a fairly high level of reliability. As sensor resolution and quality increases, this information from the third level of fingerprint detail will become more accessible. As a result, work must be done to ensure that this information can be used when it becomes widely available.; The focus of this work is to develop a biometric system that has the capability of extracting some of this highly discriminatory detail, and then using it for the purposes of verification. The system proposed in this work extracts additional detail in the form of ridge interconnection information and pore locations to augment a base-line verification system. Techniques for extracting ridge and pore information are presented, and a method for using this information to match fingerprint images is also explored. Both the multi-level and base-line systems are tested using a common database of fingerprint images collected using a high quality 500 dpi live-scan device in the UAH Integrated Biometrics Laboratory. The multi-level system exhibits a 40.64% relative reduction in the equal error rate and an 83.76% relative reduction in the false rejection rate at the zero false acceptance point when compared to the base-line system. These results demonstrate that the addition of level-three detail does indeed improve fingerprint verification accuracy.