Palmprint Identification System

Computer-aided personal recognition is becoming increasingly important in our information society. Human palmprint recognition has become an active area of research over the last decade. In this paper, a method of capturing palmprint image without any limitation is proposed, and a palmprint database has been built using this method. This palmprint database could be training set and testing set of palmprint recognition algorithm, and also could be a base of the advancement of research and development of palmprint recognition. Based on this database, we have done some preprocessing to the palmprint images. After translating the original image into binary image, we find two key positioning points in the palmprint image using automatic detecting method. The palmprint image is rotated according to the two key positioning points. Then, a sub-image of the rotated palmprint image is extracted for further abstracting of palmprint features.Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) are widely used in the field of palmprint recognition. However, the conventional PCA and LDA are both based on vectors. It means that the two-dimensional (2D) palmprint image matrices must be transformed into one-dimensional (1D) image vectors previously. The resulting image vectors of palmprint usually lead to a high dimensional image vector space. In this paper, two-dimensional PCA and LDA are used in palmprint recognition. Unlike conventional PCA and LDA that treat image as vectors, the 2D methods view an image as a matrix directly. The experimental results on our palmprint database show that two-dimensional PCA and LDA can obtain over 99% recognition rate in palmprint recognition, while using less time and memory. They are more effective than conventional PCA and LDA in terms of accuracy and efficiency.Besides, "Beijing Jiaotong University IIS Palmprint Recognition System V1.1" was built.