Fingerprint Recognition Algorithm Based On Eight Directions Gabor Filters

Fingerprint identification is based on computer recognition technology and is also the most widespread biometrics recognition technology used. The early fingerprint identification had applications mainly in forensics science. In recent years, it has gradually moved toward personal use, which resulted in high demands and needs to minimize hardware, reduce cost, increase computing time and improve recognition accuracy. The research for fingerprint recognition algorithm has invaluable and practical significance.This article mainly researches and discusses the eight directions Gabor filter fingerprint recognition algorithm in terms of the reference point localization, region of interest, normalization, eight directions Gabor filter and feature extraction cycle. The fingerprint matching is based on the Euclidean distance between the two corresponding feature vectors, which is also called FingerCodes. To improve the reference point localization, this article proposes a rotation invariance of the reference point localization method. The method solves the inaccuracy problem in localization after the rotation of fingerprint image, thereby improving the fingerprint match accuracy. Through the segmentation of the fingerprint and background, an adaptive region of interest is produced to enhance precision of the characteristic withdrawn from the fingerprint. Each individual sector of the image is normalized after the normalization process of the entire image to compensate for the change in brightness due to the difference positions and elasticity of the image. In the matching process, the new method combines the directional information with the characteristics of the fingerprint to produce a more accurate match. The improved method is of profound theoretical and practical significance because it not only take the local anomalies in the ridge structure into account but also the global pattern of ridges and furrows, inter-ridge distances, and overall patterns of ridge flow.Through experimental proof of improvement in reference point localization, ROI segmentation, normalization and matching process in established fingerprint database, the new method, having a similar computing time to the original method, greatly improved fingerprint match accuracy. At the same time, it greatly reduces the FER (False Register Rate) of the system database. In conclusion, by using a proper threshold, the method satisfies the needs of consumers.