Key Algorithm Research On Automatic Fingerprint Identification System

The Automatic Fingerprint Identification System(AFIS) is going to be in more and more people's good graces and has become one of the most important biometrics for its small size, low cost, easy manipulability and high reliability. In this paper, we focus our research on the fingerprint image segmentation, fingerprint classification and fingerprint matching, and receive the following achievements:1. Fingerprint image preprocessing algorithm is studied. We propose an adaptive fingerprint segmentation based on Harris corner detector. At first, Harris operator is used to Compute the strength of Harris corners in the fingerprint image, and then a threshold-setting method is adopted to compute adaptively the threshold for the early segmentation; After that Morphology is applied as post-processing to reduce the number of classification errors; Finally, the area parameter is used to remove remaining noised region to get the final segmented image. A thinning algorithm based on improved templates is applied in fingerprint image thinning and produces well, one pixel fingerprint image.2. The traditional fingerprint classification algorithm is studied. This paper presents a new classification algorithm based on successive pitch arc distribution of filted binarization image. We calculate the longest chord of every ridge line in horizontal and vertical direction, and the arc type of the ridge line is decided by fingerprint arc character, the fingerprint can be classified successfully by the arc types of the fingerprint ridges having.3. A new fingerprint classification algorithm based on the thinning image for live fingerprint samples is provided, it for the first time successfully extractes a new parameter suggesting the characteristics and regularity of fingerprint ridges variation by applying the ridge following method, which was defined as macro arc-like eigenvectior. The proposed method which using the new eigenvectior has been validated on the FVC2004 database and produced a classification accuracy of 98.9%. Moreover, this classification method also has a good robustness to those fingerprint images of lower quality.4. Core-based fingerprint matching algorithm is studied, and we present a new core point calculation method. First, the core region is fast located by scaning the longest chord of every ridge line, and then the core point is calculated by using Poincare index method. We define some local minutiae structures around core point, through matching these local minutiae structures, we can get some corresponding points of the two fingerprint images. Then we use the corresponding points to match the global feature of fingerprints. Finally, we promote the number of tangent lines between minutias and core points, global match distance variance, and covariance distance to help make the final decision.