Online Palmprint-based Identity Verification Technology

As one of the most important biometrics techniques, the development of palmprint recognition has significant influence on real world. The main purpose of this dissertation is to discuss palmprint. By studying the orientation, the line feature extraction, classification and the matching of palmprint, this dissertation build a system based on the online palmprint. According to the character of palmprint, the dissertation brings forward a series of methods, which are robust and fit to the analysis of palmprint. The main approaches are as follows:(1) The orientation in online palmprint system is an important and basic job. In the dissertation, we adopt an improved algorithm of line detectors to detect the contour line in the online palmprint. Then we introduce a new approach to extract feature points on the boundary. By these key points, we can establish a coordinate for the orientation.(2) Stable, simple line features of online palmprint are the key issues for popularization the palmprint identification technology. But the low contrast is conflict to the heavy noise, which can not be regularized. So how to select the proper line features is the most challenging issue. In this dissertation, a line extraction method based on cubic polynomial interpolation and minimum mean square error fitting are proposed, which can effectively extract lines out.(3) Classification of the palmprint verification is the key step to speed up the retrieval process and improve the recognition rate of the palmprint-based identification systems. In this dissertation, we compare various types of features in a palmprint and develop a classification algorithm based on structural features. This method is quite robust to rotation, translation and noise.(4) One of the key steps in online palmprint verification is the feature matching. To represent palmlines, a feature vector is defined, which contains information about the structure of the lines. Then we introduce the Fisher's linear discriminant to project palmprints from the high-dimensional line vector space toa lower dimensional feature space, in which the palmprints from the different palms can be discriminated much more efficiently.Experimental results show that the methods are effective. So palmprint identification is an interesting and important research field with theoretic and practical meaning.