Study On Key Issues In The Iris Recognition Algorithm

Biometric identification technology is based on the analysis of the characteristics of human body’s organ and the obvious difference of actions, which is a means of identification. As an important research direction of identification, iris recognition technology is uniqueness, stability, accuracy, non-invasive and wide applicability, which received more and more extensive research and related product development. However, with the complexity of iris feature and diversity of external disturbances, several key issues existing in the iris recognition technology are worth being in-depth study. This paper analyzes and investigates the issues such as iris localization, eyelid detection and iris feature extraction. At the same time, the OMAP-L138 development board and Linux-QT platform are used in this paper to demonstrate the method. The study contents and innovation were summarized as follows:(1) The detection of the inner and outer edge in iris image. In the view of the inner edge detection of iris detection, the positioning method was adopted which was based on the combination of window estimation and Hough transform to improve the inner edge localization accuracy.9 types of the scope of outer edge were determined through the analysis of the relative position of the center of inner edge and the radius of the inner edge. The precise classification localization algorithm of outer edge was given which was based on rectangular template, and the effectiveness of the algorithm was demonstrated by experiment.(2) Iris eyelid interference detection. Considering the biggest interference of eyelid in the iris image, the iris eyelid detection algorithm was given which was based on adaptive sectional line fitting. The eyelid image was processed by the combination Sobel/Canny operator and morphology. Sectional line detection was determined to do only when the iris image was covered by the eyelid or the distance between the single positioning line and the pupil was too small, and the effectiveness of the algorithm was verified through the simulation results from two aspects of the efficiency and accuracy.(3) Iris feature extraction. According to the result of the iris image detection, the iris image was normalized to rectangle image. The method that extracting feature from the radial and the angel by multi-channel Gabor filter was designed, which refines more detail feature and emphasizes the analysis and express of the localized iris texture information. At the same time, the template matching method based on hamming distance made the method have higher identification accuracy.(4) Platform implementation and verification. The ARM QT programming was used to design human-computer interaction interface. The iris recognition method was transplanted to OMAP-L138 DSP module and verified the effectiveness of the iris recognition method.