Multi-scale Smooth And Low Quality Fingerprint Singularity Detetion Based On The Three-directional Image

As a unique biological feature,fingerprint,captured by computer and subsequently recognized,can be widely used in intelligent control system such as access control and attendance check.In addition,biological research also shows that the genetic information contained in fingerprints is closely related to certain diseases.Therefore,the core task of fingerprint identification is to complete the extraction and classification of feature information.Precise ingular point location for fingerprint image is the precondition of fingerprint feature extraction and classification.Existing singularity detection methods used the low quality fingerprint images can cause the singularity position offset or location area is too large location area and thus cause unaccuracy.In order to overcome this problem,this paper has done a lot of research on the singularity detection for low-quality fingerprint images,and the main contributions are as follows:(1)Based on the calculation of the direction field of fingerprint image,the frequency estimation algorithm of gray bar graph based on direction field is proposed.The algorithm is used to estimate and construct the fingerprint frequency field by calculating the frequency with the results of the local effective ridge-valley gray bar graph.Subsequently,because low quality fingerprint image contains-noise,discontinuous,blurred or directly lost ridge line,which creates computers cannot accurately identify to singular point location.Therefore,a 2 d Gabor filter is applied for low quality fingerprint image enhancement and to effectively eliminate the influence of noise.Finally,the calculation process of the algorithm is simplified by the normalization of the fingerprint image.(2)A method for calculating the point direction of fingerprint ridge is presented.The relative position of the distribution of the pixel points of each ridge line is used to calculate its relative slope,and then the direction of the pixel points in the ridge line is obtained through the distribution of all relative slopes in the calculation field.This direction method can calculate the effective form of image texture distribution direction,and can effectively control the loss of local orientation information.Thereafter,the direction of the pixel points on the ridge line is divided into three directions,and the direction map of the fingerprint ridge line is obtained.Finally,the three directions of the fingerprint are smoothed by three directions,which can effectively eliminate the noise and complete the non-ridge area accurately.(3)This paper proposes an algorithm for detecting the singular point of fingerprint based on the three-direction fingerprint.The candidate singular points in the fingerprint can be determined by searching the intersection blocks of pixels in the three directions and the calculation of the direction sequence of the three color blocks.By using a number of different effective smooth scales can obtain more groups of candidate singular point,on the basis of detecting multiple smooth scales at the same time the singularity position and filtering fake points step by step,finally to achieve accurate positioning the singular point and eliminate the false feature points.The method is proved to be able to effectively solve the problem of inaccurate detection of singular points in low-quality fingerprint images.Based on the fingerprint image database FVC2006 testing,experimental results show that the centeral detection accuracy is 91.93%,triangle points detection accuracy is 93.36%,total singularity detection accuracy is 92.35%.The account of the singularity detection and HDSP span positioning in 10 pixels is more than 80%,which proves that the proposed algorithm can improve the accuracy of singular point detection of low quality fingerprint images,and also has high accuracy for the location of singularity point.