Palmprint Recognition Algorithm Research

The palm is area between the finger terminal to the wrist, and the palmprint is the summation of all lines on the hand including the wrinkles, the principal line and the ridges. These lines not only contain the rich texture and direction information, but also possess the characteristics of reliability, stability and uniqueness, these make palmprint become an important biological feature, which can be used for personal identification and verification. Palmprint recognition is a relatively new biometric technology. Compared to fingerprint image and the iris image, palmprint image has several advantages:low request to equipment, low-resolution imaging and more acceptable when captured etc. Because of these advantages, palmprint recognition attracted increasing researcher’s attentions, and rapidly became a new research focus in imagery processing and pattern recognition.In this dissertation, palmprint image is analyzed in detail.Based on the extensively researching the systems and algorithms for palmprint recognition, we propse two novel methods for palmprint recognition.The main contributions are listed as follows:1.To alleviate the limitation that the recent texture based algorithms for palmprint recognition lack robustness to the variations of illumination, position and orientation in capturing palmprint images, we describe a method for normalizeing the palmprint images in the illumination, position and orientation. In our algorithm, palmprint images are normalized in the orientation, position and illumination based on integrated optical density, moments and central moments.This method is more effective than the existing methods. It not only makes palmprint useful information maximization, but also to a certain extent to ensure robustness to the variations of illumination, position and orientation.2.We propose a palmprint recognition combining Gabor features and particle Swarm Optimization(PSO). In the proposed algorithm, palmprint images are normalized in illumination, position and orientation based on integrated optical density, moments and central moments, and then the normalized palmprint images are decomposed by2D Gabor filters, the means, variances and entropies are calculated for these subbands and regarded as particles to perform feature extraction by means of the particle Swarm Optimization (PSO) algorithm. A support vector machine-based classifier is employed to implement classification. Experimental results show that the proposed approach yields a better performance in terms of the correct classification percentages and relatively high robustness to the variations of orientation, position and illumination compared with the recent appearance and Gabor filters based approaches.3. We propose a palmprint recognition based on Gabor features and Pulse Coupled Neural network(PCNN). In the proposed algorithm, palmprint images are normalized in the illumination, position and orientation based on integrated optical density, moments and central moments, and then the normalized palmprint images are decomposed by2D Gabor filters. The PCNN entropy signature of the filtered images is combined with the Gabor features as the recognition features, which is classified with support vector machine(SVM). Experimental results show that the proposed approach yields a better performance in terms of the correct classification percentages and relatively high robustness to the variations of orientation, position and illumination compared with the recent appearance and Gabor filters based approaches.