Efficient Iris Recognition System based on PCA and Optimized DNN Method
Abstract
Iris Recognition System (IRS), as a developing biometric recognition method is most attractive and famous area in both practical and research uses, IR is the main procedure of identifying a user by recognizing the deceptive design of her/his iris. A representative IRS comprises the iris image acquisition, preprocessing, ROI detector, feature extraction, selection and recognition or classification. In this research work, demonstrate the existing system applied a non-traditional phase for FE (Feature Extraction), where a circular contour-let the filter bank was used to consider the iris features. This method is based on geometrical image transformation is known as CCT (Circular Contoured Transformation). The system was compared the performance with different types of iris dataset (CASIA and UBIRIS). In CASIA iris dataset performance was 96.3% and UBIRIS iris dataset performance was 96.2 % accuracy rate. The novel PALO method is selected the inner and outer quality features. The DNN classification method is used to recognize the different person samples. The experimental result defined that the iris area, diameters, peaks and feature vector for classifying the iris images according to the person identified. Overall performance has achieved an accuracy rate is 98.41%, and Time consumption is 0.198 seconds. The performance metrics are compared with existing method such as Gabor CASIA, Gabor UIRIS1, Gabor UIRIS2, GLCM, and CCT. In the Feature Extraction method, Gabor CASIA method accuracy rate is 96.30 %, Gabor UIRIS1 value is 96.20%, Gabor UIRIS 2/ Daugman value is 95.20%, Martin- GLCM value is 77.1% and CCT value is 96.30%.