The Design Of Classification Module In Iris Identification System

With the rapid development of science and technology, more and more people begin to pay attention to Identification Technology, and the traditional identification existence of security is not high , bring inconvenience and many other defects, which triggered an upsurge of biometric identification, biometric identification due its not forgotten, security features, and able to "carry" and other benefits are widely used in various fields, Iris identification is a kind of it. Iris is a protected internal organ of eye ,at the rear cornea, the front lens, which is usually the only one in the foreseeable external human internal organs. From the appearance of eyes, we know pupil at the center of the eyeball, sclera is the white part of the external, iris between of them. The surface of the iris is rugged, composes of many recess, folding and pigment spots, and contains wealth of texture information. These rich texture and structure of the iris own characteristics, makes iris is unique, high reliability advantages, as it may become a means of identification, at the same time become popular in the field of the identification of biological research.A complete iris identification system usually composition with iris image acquisition, image preprocessing, iris image feature extraction, classification and so on. The research work completed based on Japan OKI companies IRISPASSR-M iris image acquisition, IRISPASS BioAPI software development kit and iris image database experimental environment.Iris image acquisition is the first step in iris recognition, but also the most difficult step. The image quality will directly affect work of late. Due to the smaller diameter of the iris and sensitivity of eyes, making iris Image access have a lot of difficulties. In general speaking, the use of CCD camera can be difficult filming of the image, it is necessary to support infrared light for access to high-quality iris image. But by acquiring the iris image acquisition devices are usually not only covers the iris, there are often other parts of the eye, such as eyelids, eyes hair, and so on, but in a highly non-invasive system, because of the acquisition are not being asked , in the image of the iris position and size will change. Therefore,First determined the location of iris in the image and the size of normalization before iris recognition. In some cases, the iris image illumination is uneven, so that will impact the accuracy of iris recognition. At the same time, iris, the pupil size of the border that will change, which make the iris texture will deformation. These conditions will affect the quality of the iris image, making the next step feature extraction and accurate matching difficult. In order to achieve accurate matching, it is necessary to eliminate the above factors through the pretreatment.Paper's proposed two-step iris location ideological, the rough edge of the iris image location use of histogram determine pupil separation threshold, binarization and then geometric projection; On this basis, Sobel operator with Hongh transform the edge of the precise positioning of it; After histogram equalization enhance image contrast by calculating the maximum gray level changes regressive value-added to the rough edge of the iris position; Using of improved Daugman operator precise positioning edge of iris; The algorithm make search for iris in the image more easy and shorten the calculation time and improve the positioning accuracy.After image pretreatment we need analysis of the data, looking for the details, this process is named iris image feature selection and extraction. Because the original image data is so large, we need to convert it to a number of characteristics, and this is feature extraction. In order to enhance separation of the speed and accuracy, the extracted features also need to select the most representative characteristics, which have the smallest of information redundancy. The applications Classifier of iris recognition technology general classified into two categories, one is Verification, the system using to be classified iris coding matching with one of particular iris coding template in the database, namely, One-to-One Matching, the results will be output" target and template whether the same person". Another is Identification. In this mode, the system using to be classified iris coding matching with all of the iris coding template in the database, this is One-to-Many Matching, Recognition results will be exported this person whether or not existing ,if exist who is he/she.No matter what kind of question is can be attributed to statistical decision theory framework. Through this method we can also transformed pattern recognition problem into a more efficient for the task, it requires implement a Statistical Independence test.As to design to classification module, this paper first does research taking Hamming distance classifier as the start point. According to statistics to a great number of training samples, using particle swarm optimization(PSO)algorithm to get the minimum error ratio and classification threshold of classifier. The result get from theory has been validated according to huge number of sampling testing. For the sake of shorten classification time of iris identification system,"rough classification"has been introduced at the basis of Hamming distance classifier. After quadratic classification like this, if some of the critic features have not been matched, then there is no need to match all the other iris features. By that, the identify work load has been reduced, classification time of classifier has been shorten, the system identify speed has been improved in some certain degree.