Research On Palmprint Recognition System Arithmetic

System and information security is becoming increasingly important in the information society. Personal authentication is becoming necessary in more and more field. Personal authentication methods is a new fashion of the development of society. Under such circumstance, there has been a high demand for Biometrics for security reasons.Biometrics are the technologies that analyze human characteristics for automated personal authentication. It is concerned with the unique, reliable and stable personal physiological characteristics such as fingerprints, palmprint, facial features, iris pattern, retina and hand geometry, or some aspect of behavior, like speech and handwriting etc. Research on the issue of fingerprint identification and speech recognition has drawn considerable attention over the last 25 years. Recently, issues on face recognition and iris-based verification have been studied extensively, which results in successful development of biometric systems for commercial applications. However, limited work has been reported on palmprint identification and verification.Palm is the inner surface of a hand between the wrist and the fingers. Palmprint is referred to principal lines, wrinkles and ridges on the palm. Like fingerprints, palmprint has been used as a powerful means in law enforcement for criminal identification because of its stability and uniqueness.As an important complement of the available biometric features, palmprints have their own virtue: Normally, people do not feel uneasy to have their fingerprints taken for testing. People can not accept the recognition testing. But if hand images and prints be taken for testing, people will feel easy. so, palmprints can be accepted by people than fingerprints; palmprint capture devices are much cheaper than iris devices; the features of palmprints are much more stable than those of signatures; palmprint recognition can obtained higher accuracies than face recognition; palmprints contain additional distinctive features such as principal lines and wrinkles, which can be extracted from lower solution images; and a highly accurate biometrics system can be builded by combining different features of palms. It is for these reasons that palmprint recognition is becoming a hotspot in the biometric filed after it appeared in last several years.However, how to develop an automated recognition system for identification and verification of a large collection of palmprint image data with accuracy and reliability remains a challenging task. in general, the design of an automated biometric recognition system involves image data collection, image preprocessing, feature extraction, feature matching.This paper concentrates on the research of the key technologies and algorithms in palmprint identification system. According to the detail research of other biometrics, and considering the characteristics of palmprints, we propose a series of efficient palmprint preprocessing, feature extraction, feature matching algorithms. In detail, this paper concentrates on the research of the following key technologies and algorithms:1.Palmprint preprocessing, alignment, and segmentation:Generally, noiseAlways exist in palmprint images. Besides, different palms have different sizes,and because of the translational and rotational placements of hands, thecollected palmprint images from the same palms are different from time totime. Therefore, it is essential to process the palmprint images includingalignment, normalization and segmentation before feature extraction.This paper proposes a new method to detect the contour points of palmprint, based on the property of Harris corner that the points'area (such as the area of corner, the area of outline etc.) have some relation with their first derivative. The paper make use of this property, combine with the physiology characteristic of palmprint's contour points (convex and cave) to design the detect algorithm. Get the potential points, then categorize these potential points according to the finger width, the maximal value of y will be the optimum point. By these optimum points, we can establish a coordinate for the normalization. The experiments show that this method has low computational complexity, high accuracy.2.Feature extraction and feature representation:Analyze the palmprint textures with a multi-resolution method. Palmprints can be regarded as textures composed of some basic elements such as palm-lines with different width and ridges. These basic elements have inherent different resolutions, therefore they should be analyzed with multi-resolution ways: the thickest and widest palm-lines can be analyzed in low resolutions; the medium thick and wide palm-lines should be analyzed in medium resolutions and the thinnest ridges must be analyzed in high resolutions.This paper taking advantage of the multi-resolution of the textured image, we obtain the multi-resolution analysis of the textured image with wavelet transformation. Then we do the wavelet transformation and extract feature vectors, then we use the energy and variance in LL, LH ,HL and HH as the vectors. We defines the vectors as a novel palmprint feature, which called wavelet texture feature. It can reflect the wavelet textures distribution of palmprint in different directions at different resolutions (scales), thus it can efficiently characterize palmprints3.Features Classification and recognition:Classification is the key step to speed up the retrieval process and improve the recognition rate. Compared with other various features in palmprint, a classification arithmetic based on texture features is developed and it is quite robust to rotation, translation and noise.After obtained the texture feature, we use ISODATA clustering method to achieve the classification of the Feature vectors. ISODATA clustering method can adjust the classification of every Feature vector during the iterative process. The number of the clustering can also be adjusted by "splitting" and "uniting". Thus we can achieve unsupervised texture clustering for unknown texture image. We call it is coarse classification. Then, based on the coarse classification, we can calculate the Euclidean distance from one feature to another in the same class. Therefore, we can finish the recognition of the original image by the minimum of distance. Thus we achieve fine classification for the feature.Experimental results based on large data sets illustrate the good robust and effectiveness of the proposed approach for palmprint recognition.