The Research Of Some Key Technologies In Automatic Fingerprint Identification System

Fingerprint Identification has wide application prospect in all fields which contain identity authentication. Constructing reliable, safe and practical operation of the automatic fingerprint identification system (AFIS) has become a research hotspot. Although theoretical research and application development of AFIS has made a significant progress, accuracy of the algorithm and processing speed still need to be improved. In this paper, fingerprint image preprocessing algorithms, fingerprint singular points and minutiae extraction algorithms and fingerprint classification and indexing algorithms are analyzed and discussed in detail.Productions are enumerated as follows:1. In the aspect of fingerprint image preprocessing algorithm, firstly, feature vectors composed of Fourier spectrum energy ratio and gray contrast are extracted from fingerprint images sub-blocks and classified by linear support vector machine. After classification, morphological operations are performed to ultimately realize fingerprint image segmentation. Secondly, an improved fingerprint image enhancement algorithm based on Gabor filtering principle is presented, which uses adaptive smoothening to estimate fingerprint orientation field and adopts spectrum analysis of projection signal to estimate ridge distance.2. In the aspect of fingerprint feature extraction algorithm, firstly, an improved method of fingerprint singular points extraction is proposed, which estimates complex square pixel-wise orientation field by multiple scales low-pass filtering and takes three kinds of complex filters on the orientation field. Then, more accurate singular points' location and orientation are determined according to the response amplitude and phase. Secondly, an improved minutiae extraction algorithm based on fingerprint thinned image is presented, which uses open, close and fill operations in fingerprint binary image to avoid producing complicated false minutiae structure and removes false minutiae in the foreground edge according to segmentation mask, and then integrates several kinds of post-processing rules to eliminate different types of false minutiae in fingerprint thinned image.3. In the aspect of fingerprint classification and indexing algorithm, firstly, a fingerprint indexing algorithm based on average ridge distance is presented, which performs continuous search in the template database using average ridge distance, meanwhile arbitrary classification precision can be realized by adjusting search radius. Secondly, a fingerprint indexing method based on singular points' orientation is presented. The first classification is made according to the singular points' number and their relative positions. Fingerprint indexing is implemented through contrasting feature vectors which are composed of the orientation field around the singular points.