On-line Palmprint Matching Technology Research

Palmprint recognition is the method of identification through the palm veins characteristic, which is an innovation of the traditional identification methods and has already become one hot field of identification. At present, the palmprint recognition technology mostly focused on structural features, ridge characteristics and geometric features of the palmprint images, and the study of the structural features and ridge characteristics is the key. Because the Palmprint image exist a variety of interferences from the external environment in the collection process, and the application of processing technology is not mature, the palmprint recognition research and development are greatly restricted and it is impossible to satisfy the real-time requirements of palmprint identification system. According to these characteristics of the analysis of palmprint recognition and on the basis of analyzing the key technology of palmprint recognition, the researches on a regional location and feature extraction are respectively implemented.In order to improve the characteristics of palmprint identification match rate, through the analysis and research of the existing palm matching algorithm, the extraction algorithm for palm corner point and the ROI region has been improved and the corner detection algorithm based on the edge curvature of the palm is presented. In addition, the corner points extracted is used to establish a rectangular coordinate system and the ROI region selection of palm image is achieved. It greatly reduces the matching algorithms affected by the rotation, twisting and translational nature and effectively improves the match rate.The improvement from the feature extraction is implemented so as to further reduce the disruption of the palmprint images during the collection process. The advantages of directionality and anisotropy and the characteristic of the translation, rotation, scale and contrast transformation invariant features of contourlet transform are used, and an algorithm of adaptive weighted palmprint recognition combining the màtrous-contourlet transform with invariant moments is presented. At the same time, the characteristics of decomposition coefficients in all directions coefficient are extracted as a weighting coefficient by means of the frequency-domain transform characteristics which are similar to gauss transformation graphic. The mode of"even better"is used to achieve palmprint image matching and recognition and the problems of various translations interferences are solved in the process of the palm collection. The experimental results show that the positioning time improves 0.4s more than the traditional algorithm and it raises nearly 10% on condition that the accuracy rate is as high as 99.3%. The accuracy and real-time of the online palmprint detection are effectively improved.