The Research Of Improved Pre-processing And Classificaton Method On ICA/PCA Iris Recognition Method

In this paper, setting on practical environmental requirements researches for iris recognition algorithm.To improve pre-processing and classification of the iris recognition, we achieve the ICA/PCA iris recognition method which is normalized means on ROI and used the bionic thought to classify, and in the implementation process of the algorithm, we complete relevant work.In the thesis positioning stage, I study the related literature at home and abroad, and illustrate the advantages which compare with the biometric technology and the traditional recognition technology.Introducing the research background, it compares various aspects to the iris, voice, fingerprint, face, signature, and several common gait and other biometric recognition technology, including the security, collection, unique, universal, stability, receptive and anti-deceptive. Furthermore, it illustrates the advantages of iris recognition and the importance of this research.Overview of stages in the iris recognition, from the general direction, it divides the iris recognition into registration and identification of the two processes. From the small direction, it divides the iris recognition process into various stages.It includes the iris image acquisition, quality assessment, iris localization, normalization, iris image enhancement, feature extraction, structural characteristics and the pattern matching etc.Each stage of the iris recognition process makes the situation analysis in domestic and foreign, thereby it's original to illustrate adopting the iris identification method in the paper.The paper introduced the laboratory with independent intellectual property rights of the iris image capture device which includes meanly the appearance and components of the acquisition device (goggles, lighting systems, data transmission system and imaging systems) and the parameters of the various components. Finally, it provides the iris image collected using the acquisition device to.Construction of the iris image database on the practical environment, it is divided into different periods, different environment for the iris image acquisition. Using the appropriate algorithm cuts frame and filters image, moreover, through the certain organization, we complete this construction of the iris image database.The process uses iris image quality assessment algorithm detecting the iris spot sequence, the integrity of the iris region, image brightness and living image, and filtering image gradually.Preserved image into the airspace image, interference factor calculated and resolution factor gets comprehensive quality evaluation factor. Ultimately retained the image can be used to identify.It introduces the iris recognition process at all stages of pre-processing, including through Canny operator for contour extraction and combining with Hough transform for iris localization algorithm to locate, cylindrical localization algorithm on single maximum gradient, the iris normalization method of polar coordinate transformation and bilinear interpolation on tradition and the iris normalized method on ROI, the iris images enhancement method on local histogram equalization.It emphasizes on comparing the iris normalized method on ROI to the advantages of the traditional normalized method.This paper describes in all aspects of the algorithm of the ICA/PCA iris feature extraction, illustrating the three aspects of the ICA algorithm, PCA algorithm and improved algorithm for PCA on the ICA, and introducing some concepts and independence of the criterion with ICA/PCA algorithm related. Moreover, ROI-based means of normalization applied to the pre-processing stage and implements the ICA/PCA iris recognition method on ROI normalized means.It leads the idea of bionics into the method of the ICA/PCA iris recognition, and it improves the classification of iris recognition. From multiple angles, compared the biomimetic pattern recognition with the traditional pattern recognition of different points, It indicates that the biomimetic pattern recognition process is more similar to the process of human cognition.To illustrate the specific process of bionic category, it introduces some relevant definitions of high dimensional space(Including the distance between two points, angle, super flat, super-sphere and the sphere), PHC principles and the definition of coverage, Topological model and the definitions of the optimal coverage etc.It introduces the classification method based on partition of traditional high-dimensional space and the classification method biomimetic pattern recognition based on cognitive, and gives the simplest super-sausage Topology model.To introduce topology model construction method on ICA/PCA algorithm. At first, it introduces the cosine distance classifier and three-dimensional model.It illustrates the source of ultra-vertebral biomimetic topology model in the paper, the ultra- vertebral topology model of the specific algorithm and uses maximum likelihood estimation method for threshold selection.In the recognition process, it compares the angle similarity of the feature vectors of the test sample and the primitive axis.To verify the effectiveness of the method of iris recognition, we carry out related experiments. It includes mainly ICA/PCA iris recognition about the sensitivity testing experiment of iris image rotation,which illustrates the rationality of the ROI selection,the basis of the experimental of directing PCA principal component vector to select number,the selection experiment of ROI size and location of the horizontal iris, the selection experiment of ROI size and location of the vertical iris, and comparative experiment of different identification methods. It verifies finally the effectiveness of the iris recognition algorithms.