FINGERPRINT MATCHING USING ZERNIKE MOMENTS ON CONCENTRIC CIRCLES AROUND CORE POINT

Abstract

Accurate and reliable automatic personal identification is critical in wide range of application domains such as National ID card, Electronic Commerce, ATMs etc. Biometrics which refer to automatic identification of a person based on his physiological or behavioral characteristics is inherently more reliable in differentiating an authorized person from an imposter, than traditional password and PIN number based methods. Among all the biometric techniques, fingerprint based authentication is mostly used because of its reliability, low cost and ease of integration. Fingerprint indexing is an efficient technique that greatly improves the performance of Automated Fingerprint Identification Systems. Continuous fingerprint indexing method based on location, direction estimation and correlation of fingerprint singular points has been analyzed in detail. There have been many approaches introduced in the design of feature extraction. Based on orientation field, firstly, it is divided into blocks to compute the Poincare Index. Secondly, the blocks which may have singularities are detected in the block images. For fingerprint matching, an approach based on localizing the matching regions has been proposed. The location of region of interest is determined using only the information related to core points based on feature vectors extracted for each fingerprint image by Zernike moment invariant. Zernike moment is selected as feature descriptor due to its robustness to image noise, geometrical invariants and orthogonal property. Using the singular points, the area around the core point has been cropped into four concentric circles and Zernike moment is applied on each of them. To find out the matching difference among Zernike moment invariant feature, normalized Euclidean distance is calculated among the two corresponding Zernike moments invariant features, stored template and query fingerprint image. This idea is applied on FVC 2002 Database which consists of 100 classes, each class having 4 training and 4 testing images. The parameters used to compute the performance are false acceptance rate and false rejection rate. A genuine match is done by matching a testing image of a v class to a training image of the same class, whereas for an imposter match is done by matching the testing image of a class to the training image of another class. To calculate the Equal Error rate Zernike moments orders were varied from 0 to 15. By increasing the moment order the EER started to deteriorate, but at order 13 and onwards the results started to converge and EER started to increase rather than decrease. So the best moment order selected for this approach was 12 which resulted in giving a minimum error rate of 16.59%. This results in a recognition rate of 83.41% of the proposed system.