Research On The Key Technology Of Iris Recognition

With the development of information technology in contemporary society, traditional identification methods cannot meet the need of information security. Automated personal identification based on biometrics has been receiving extensive attention. Compared with other biometrics, iris identification has more advantages in uniqueness, stability, non-invasion and anti-falsifaication. In recent years, iris recognition technology has made great progress, and its applications have wide development prospects in many fields.Iris recognition system consists of iris image acquisition,image preprocessing,feature extraction and pattern matching. The thesis focuses on the key technologies of iris recognition: iris location, feature extraction and pattern matching. The main contributions of my work are as follows:Firstly, an improved method is proposed to make up for the deficiency of classic location algorithms, it accomplishes the location of inner boundary and outer boundary using the gray distribution features of eye images. Experimental results show that the algorithm proposed is more accurate and faster.Secondly, a feature extraction method is proposed based on texture feature of the partial area of iris image, only the area which includes abundant feature information is selected. Haar wavelet packet analysis is performed on the feature extraction, leaving only the low frequency and vertical low frequency components of the feature vector to encode. The encoded data is greatly reduced, the speed of feature extraction improved significantly.Thirdly, the cyclic shift algorithm is used to solve the matching distortion caused by eye rotation. The algorithm uses the Hamming distance classifier, searching the minimum distance of iris codes as the basis for the final match after cyclic shift. The optimal shift times of matching algorithm is determined through experimental analysis.All the above algorithms are simulated on the platform of Matlab 7.1 with the CASIA V1.0 iris database. Experimental results show that the iris recognition algorithm proposed is effective and satisfactory in improving iris recognition accuracy and speed.