Research On Modular Palmprint Recognition Algorithms

With the development of technology and social informatization, biometricidentification technology has enjoyed a quick development. The fingerprint、face,iris, palmprint, hand, shape and other characteristics of human can be used as thebasis for authentication. Palmprint identification is a kind of biometri cidentification technology, which has received considerable attention in recentyears. With its many merits including rich information of palmprint image,simple acquisition equipment and high accuracy and so on, palmprint recognitiongradually gained more and more attention. In recent years, palmprintidentification system has appeared in market.This paper focuses on palmprint feature extraction, especially the palmprintfeature extraction based on module methods, which is the most critical aspect ofa palmprint identification system. The main research work of this paper is asfollows:The palmprint recognition algorithms based on PCA are investigated indetail. This paper details the implementation steps of PCA, module PCA,2DPCA,module2DPCA,(2D)2PCA algorithms, and through a large number ofcomparative tests, the correctness that module method can improve theeffectiveness of palmprint recognition rate is verified. Module method can notonly reduce the dimension of palmprint image, which can increase the number ofsamples, it can also extract local features which are beneficial for classification.As traditional modular method does not take the difference among blocksinto account, this paper makes an improvement on the traditional modular basedprinciple component analysis algorithm. At first, take the sub-blocks which havethe same location of all the palmprint images in training set as a child training set,then compute their scatter matrix respectively. At last, take traditional modular(2D)2PCA algorithm as an example, the good performance of this new method isverified.The palmprint recognition algorithm based on modular LBP is investigatedin detail. The problems of too long recognition time and too much storage spacewhich are aroused by high feature dimension are briefly described. On this basis, the LBP+PCA palmprint recognition algorithm is proposed. In this algorithm,PCA is used to reduce the dimension that LBP extracts. The experimental resultsshow that this method can significantly reduce the dimensionality of theextracted features without affecting recognition rate.