| In this paper, research has done on the key technology of iris recognition, after which the iris image quality evaluation is raised as well as a rapid and effective iris boundary localization algorithm where a kind of method based on line extraction is used for iris normalization based on image block is put forward. In terms of pattern matching, a method of improved Hamming distance calculation combined with cyclic shift has been used; finally, simulation experiment was carried out in the iris database. The experimental results show that the improvements have obvious advantages compared with the conventional algorithm:First, the iris positioning time is shorter; second, the iris features extracting is more completely, and third the matching results are more accurate.The following aspects are mainly worked in this paper:(1) An iris image quality evaluation link is studied. Before the iris feature extraction carried out, some unqualified image can be filtered through the comprehensive quality evaluation factors obtained by the calculation of the resolution factor and the disturbance impact factor of the iris image, as well as the fake iris filtered through the live diction and the Fourier transform of the iris in the security test.(2) The iris localization algorithm is improved. After the iris image gray histogram selected, appropriate threshold to isolate the pupil from the iris is chosen through the algorithm, and then the center of the pupil through the method of Gray-level projection geometry is located. When locating the inner boundary, firstly, prior knowledge is used to determine the image block of locating inner boundary based on pupil course positioning, where the points which is away from the boundary is filtering out to reserve the inner boundary area only, and the Hough transform is used to search inner boundary parameters within the image block aiming at finding out iris inner boundary accurately. Similarly, when positioning the outer boundary, the points which is in the outer radius or in the inner radius is filtered by using the prior knowledge, a smaller image block got, and then the Hough transform is used to locate iris outer boundary accurately. It is conductive to improve the positioning accuracy and the positioning speed of the iris by the use of the method combined by the coarse positioning and fine positioning, through which the point nothing to do with the boundary location can be filtered and then the search range of parameters made smaller.(3) The iris normalization processing is improved. Considering that there are some limitations in the Daugman rubber sheet model, a more line segments extraction method is proposed for the iris normalization processing, where the main principle is that a circular iris region divided into n parts averagely and then the multiple lines is used to describe regions between two different heart round.(4) The iris feature extraction is optimized. An iris feature extraction algorithm based on Log-Gabor filter is used to extract iris texture feature, which method not only possesses the advantages of the Gabor filter, but also can construct filter of any bandwidth, and won’t produce DC component. So the characteristic value of the iris can be more fully extracted.(5) The pattern matching is perfected. In this article, improved Hamming distance is brought out to calculate and cyclic shift the coding method to match the iris. Built a interface template Q by the interface points such as eyelash and eyelid, the Hamming distance can be improved, meanwhile, the position which has the maximum similarity is found as the Hamming distance matching with the iris figure. |
PDF Full Text Request