The Research Of Iris Recognition Technology Based On Scale Invariant Feature Transform

Iris recognition technology is a unique biometric identification technology and used in the positioning of iris texture information,by virtue of its uniqueness,high stability,high security,non-contact,non-invasive and in vivo detection has become a hot research field of bio recognition.It has a broad market prospect,scientific research value and huge economic benefits.The process of iris recognition includes four parts: iris image acquisition,iris image preprocessing,feature extraction and feature matching.Iris image preprocessing includes iris localization,normalization,image enhancement and image denoised.In this paper,the following aspects are studied:Iris localization is divided into inner and outer boundaries.Firstly,the localization of the iris inner boundary is solved by the method of dilation and erosion,while the pupil area projected in the vertical direction and in the horizontal direction is used to locate the center of the pupil.Secondly,using the feature of the most obvious change of the gray value of the pupil boundary,we can find the three points on the different line.The inner boundary of iris is precisely located by using the decision theorem of circle and the radius of pupil is found out.The location of the outer boundary is based on the Canny edge detection operator and the Daugmen detection algorithm,then the outer boundary of the iris located.Because of the diameter of the iris is about 5 times of the pupil diameter,the localization of the iris outer boundary can be searched within the range of the pupil radius of 2.5.Using the Scale Invariant Feature Transform(SIFT)algorithm to extract the texture feature vector from the processed image directly,which has good characteristics of constant scale,rotation,illumination,noise and avoid the normalization process.Because the extracted feature vector is 128 dimensional,the memory is very large,and the extracted feature points are easily affected by the eyelashes and other factors,which will affect the subsequent feature matching in a certain extent.So,in this paper,we propose the Harris corner detection operator to select the initial feature points,and then use some high contrast and high qualityfeature points as the final feature vector.In the aspect of feature matching,this paper uses block distance to match features.Compared with the Euclidean distance matching algorithm of the original SIFT,the block matching algorithm is used to ensure the precision constant while can improve the matching time.In the experimental results,the 20*8 image is selected as the training sample from the CASIA-V1.0 iris database.The experimental results show that the improved SIFT algorithm has a better recognition rate compared with the original algorithm.In order to further explain the universality of the experimental results in this paper,we use CASIA-V1.0 and CASIA-V2.0 two databases as experimental samples to compare the original SIFT algorithm and the improved SIFT algorithm.Experimental data show that using the Harris corner detection algorithm on the original SIFT algorithm to extract the feature points after filtered by using the method of block matching distance in guarantee the algorithm robustness can make the correct matching rate has greatly improved,and reduced the error matching rate,to a certain extent,which can improve the efficiency of the operation of the algorithm,and provide guarantee for some quick applications.