Design Of Iris Recognition Algorithm And Improvements Of System Software And Hardware

Biometric recognition is a technique which can automatically perform personalidentification/verification based on physiological or behavioral characteristics.Physiological characteristic, such as Iris, fingerprint, face, voice, palm print etc, areinnate. However, behavioral characteristics which are the habits of human, such ashandwriting, gait, are postnatal. Compared with traditional personal recognitionmethods, biometrics has some outstanding advantages, such as don’t have toremember intentionally, not easily be forged, available anytime anywhere. Irisrecognition which is proved to be one of the most reliable biometrics, is performedmainly based on the intricate structure with many minute characteristics such asfreckles, furrows, crypts, wrinkles and triples. This dissertation will elaborate how toimprove the system’s hardware and software technology in order to improve theperformance of the recognition system. Our contributions are as follows:（1）A new electronic control module and a auto focus program are designed inthis paper. A new MSP430F149based electronic control unit of iris acquisitionsystem is designed, which uses light source of810nm to enhance the performance ofcamera. This system can assist users’ fast self-positioning and compensate thefluctuation of ambient light. It detects the users and the distance between the capturedevice to feedback position signal to the zoom lens. Voice prompt signals are used toimprove system’s human-computer interaction in image acquisition, registration andidentification stage. An auto-focus lens control method based on Laplacian ofGuassian is implemented to improve the success rate of acquisition and focus speed.（2）Research on the image preprocessing and quality evaluation. Two discretewavelet transform based quality indicators of clarity and blocking factors areproposed to select iris images which are clear enough and not covered by eyelid andeyelash. Ellipse fitting is used for accurate positioning after rough location of thepupil. And then, fast position of the outer edge of iris is realized by the method ofdetecting edge gradient using3×3neighborhood. The detection techniques ofeyelids, eyelash and specular reflections are proposed in this paper. In addition, local histogram equalization method is used to eliminate effects of uneven illumination andenhance the contrast of iris texture.（3）This paper discusses how to optimize parameters of the feature extractionand code matching. It reveals the rules between discriminating index and Gabor filterparameters, such as frequency, direction, scale factor. In addition, it analyzesparameters effect on the performance of iris recognition system and proposes a newmethod to design the two-dimensional Gabor filters parameters. In addition, itdiscusses the principle of local binary pattern to extract feature and how to use thismethod to express iris feature. The experiment result shows that the systemperformance is greatly improved by the discriminating index from5.64to7.64.（4）Establish the iris image database. We used our self-made iris image capturedevice to collect500people left and right iris image. An iris image qualityassessment and file operating platform is established to realize read and write, storage,evaluation and classification of batch image data.6000iris image were selected fromthe original data to constitute the iris database THP-Iris V1.0which involves5sub-databases, such as the database of clear image, clear image covered by eyelidsand eyelashes, iris image of born defects, iris image of distortion and thesub-database of blurred iris image. They can meet the training and testingrequirements for different algorithms.