Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study on Automatic Fingerprint Identification Technique

Author WangChongWen
Tutor YuanXiangHui;LiJianWei
School Chongqing University
Course Instrument Science and Technology
Keywords Fingerprint Matching Fingerprint Classification Detail feature Integrated template Image preprocessing
CLC TP391.41
Type PhD thesis
Year 2002
Downloads 1421
Quotes 37
Download Dissertation

Traditional security system is mainly based on the token or password With the development of society, such a system is becoming more and more fragile. To meet these challenges, the people turned to biometric technology, hoping to help the body's physiological characteristics or behavior of action to improve the accuracy of the identification. Will produce a variety of lines in the surface of the finger, because uneven skin, often these lines appellation fingerprint. Modern anatomy and statistics have shown that each person's fingerprints are lifelong constant and not with other people, at the same time, everyone fingers lines vary. That is, the fingerprint stability and uniqueness. Fingerprint as one of the main biometric technology is widely used in the administration of justice, public security, and a variety of security protection system. Fingerprint recognition as one of the first to set foot in the field of pattern recognition, the extremely rapid development, has become synonymous with the biometric technology in fact on many occasions. Automated Fingerprint Identification is a comprehensive application of image processing technology, the technology of pattern recognition technology and a computer database. One of its research can be traced back to the late 60s of the last century, after nearly 40 years of development in the field of justice, public security, and some commercial has been made in the successful application of hot spots, but the current AFIS pattern recognition research. Article in the collection and analysis of a large number of home and abroad in recent years on the fingerprint matching and fingerprint classification academic papers, research reports, a number of theoretical issues of the computer fingerprint recognition technology discussed three main technical aspects of automated fingerprint identification method - - of fingerprint pretreatment, fingerprint matching and fingerprint classification conducted in-depth research, proposed a set of fingerprint pretreatment of the crystal fingerprint scanner, fingerprint matching and fingerprint classification method, the test to verify the present fingerprint pretreatment fingerprint matching and fingerprint classification method is reasonable, the theoretical value and practical value. The contents of this article is organized as follows: The first chapter describes some of the commonly used biometric technology, such as fingerprint, face, iris, hand geometry, signature, voiceprint, etc., and from stability, uniqueness, accuracy, acceptability etc. seven aspects of these technologies. At the same time, it also introduced a biometric system evaluation parameters, as well as two commonly used operating mode. While fingerprint identification has been some criticism, but as a proven scientific reliability of fingerprint identification is indisputable. Next, the article describes the structure of the Automated Fingerprint Identification System, a detailed analysis of fingerprint classification, fingerprint matching and fingerprint compression functions and composition of each subsystem. In the end of this chapter, pointed out the problems faced by the current Automated Fingerprint Identification System fingerprint acquisition, image enhancement, feature representation and classification system, and the countermeasures. The second chapter describes the pre-processing of fingerprint images collected using a crystal sensor. First of image quality assessment, this paper presents a new fast fingerprint quality assessment algorithm, by the pattern of the fingerprint image of low sampling to determine the fingerprint fingerprints taken is too small, or the fingers put too biased, or too wet finger or too dry. Followed by image segmentation, the integrated amount of segmentation feature using hierarchical segmentation system, to ensure the speed and accuracy of the segmentation; then the image segmentation the recoverability fuzzy region using a fast Fourier transform into lt; WP = gt ; line enhancements, and the enhanced image separation and the average filter to further eliminate the ridge line between the fork connected and fracture; enhanced grayscale image fed to the binarization module processing, as used herein based on the direction information of the new binarization methods, the use of image and binary the direction weighting filter denoising, achieved good results; fingerprint thinning quality is the key to extracting fingerprint features, this paper, the classic refinement algorithm based on the use of a kinds of nearest neighbor extract ridge skeleton, meet the fingerprint thinning to maintain connectivity, the central axis and fast requirements. The third chapter studies how to extract the the the fingerprint template using the integrated template fingerprint matching method. Called templates, that take into account not only the minutiae location, direction, and type, but also consider local texture characteristics and confidence. Therefore, we first describe the thinned fingerprint image pretreated extract detailed features and characteristics according to the details of the topology and statistical distribution law, the detailed features extracted pruning to retain the most credible details. Taking into account whether it is generated fingerprint registration template matches are an indispensable part of the details of the point-based mode or in the process of fingerprint matching, we introduced a fast matching algorithm based point mode details. The method uses the idea of ??clustering to first identify the matching points with maximum support number to calibrate two points under the point of set to draw point set match score. Finally, we proposed the details coding (MinutiaeCode), thinking. The encoded record only the details around the point of having a large number of identification information of the local texture features, and in accordance with the details of the encoding and details template matching ultimately generate consolidated template. The integrated template fingerprint matching is actually a two-tier system, the first match to detail characteristic, if unsuccessful, then use the details of coding, the final match score to determine both the results of the match. In short, whether it is to generate comprehensive template, or use of the integrated template matching, we have adopted the strategy to improve system performance information fusion and information fusion impact on the overall performance of the system is analyzed theoretically. The experimental results show that the proposed integrated template-based fingerprint matching method has certain advantages relative to several other methods in both accuracy and efficiency. The Fingerprint Classification is Chapter IV. The fingerprint classification is an important part of the identification system, the traditional fingerprint classification method basically mimic artificial classification based on the fingerprint classification method proposed in this paper: the hidden Markov support vector machine-based two-tier classification using fingerprint encoding . The method uses FingerCode expressed as the characteristics of the fingerprint, the first five pseudo 2D hidden Markov model category primaries, to determine the most likely of the two fingerprint classification results to make the final judgment, then the corresponding support vector machine classifier. At the end of this article, we also use the space distribution of the fingerprint encoded, trying to establish a new classification system: the use of fuzzy clustering method re-clustering in the feature space. The experiments show that the classification accuracy to meet the basic needs of general application. At the end of the paper, we give a summary of the full text, and pointed out the direction of further research.

Related Dissertations
More Dissertations