Contactless Fingerprint Recognition Algorithm Research

At present, with continuous development of computer technology, biometric identification technology has made unprecedented progress. Because of characteristics of uniqueness and permanent, fingerprint identification technology has become preeminent in the field of biometric identification technology and become the most commonly used means of identification.Nowadays, automatic fingerprint identification system is generally realized in contact way in society. This way of acquisition of fingerprint image is easily affected by environment,such as deformation due to pressure problem or false image with reminders on fingerprint sensor and so on. These problems will make fingerprint matching failed. In order to fundamentally solve these problems, non-contact acquisition method was brought forth. In this article, non-contact fingerprint image acquisition and the relevant algorithms are researched.As non-contact acquisition of fingerprint image has certain quality problems, such as low contrast, low definition, we put forward corresponding image preprocessing, feature extraction and matching algorithms. These designed preprocessing algorithms include curvelet denoising, enhanced local adaptive binarization and Gabor. Curvelet transform is a bandpass, multi-resolution and directivity function analysis method, which can highlight the edge curve and enhance the texture of interval, especially suitable for fingerprint image. As the problem of low contrast, where gray value contrast of the valley of ridge line is not strong,even the grey value overlapped, curvelet transform is also used to effectively solve it. For contactless fingerprint image processing, in this paper, the global threshold will not be able to obtain images of the ridge line and line separating, so we adopted the local adaptive binarization method. The method for each widget adopted OSTU algorithm to determine a threshold, and finally realize significant separation of ridge and valley line in fingerprint image. Because the images are still have some problems such as fingerprint ridge breaking and the noise, in this paper, Gabor enhancement algorithm is adopted to compensate for fracture and remove noise. The biggest characteristic of this method is a directional characteristic and frequency selective characteristic. To fingerprint image with strong texture features, Gabor filter can effectively supplement breaks in direction field and enhance the intervals in frequency field.There are many features in fingerprint image, the most representative features are detail features, including fingerprint endpoints, bifurcation points, center point and delta points, and so on. Since the extracted fingerprint region is relative small, only the end points and bifurcation points are used as feature points. In the matching process, the first step is fingerprint image registration, then use boundaries-box approach to do image matching,through experiment,the matching rate and matching speed has reached a basic requirement.