Online Palmprint Image Analysis And Recognition

As one of the most important biometrics techniques, the researches of palmprint recognition have significant influence on real world. According to the character of online palmprint, the dissertation brings forward a method of palmprint recognition analysis. The dissertation deeply studies the orientation, the palm-lines enhancement, the palmprint features extraction in the spatial domain and frequency domain, the matching of palmprint features. The experimental results demonstrate the feasibility of the proposed method. The main approaches are as follows.The orientation in online palmprint system is an important and basic job. Based on the character of online palmprint image, the dissertation extracts the contour line of the palm using mathematic morphological algorithms. Then we introduce a feature point extraction approach to extract feature points on the boundary. By these key points, we can establish a coordinate for the orientation and segmentation.A method is presented to enhance palmprint feature lines with Gabor-filter. In the algorithm, the palmprint is divided into many sub-blocks and then these sub-blocks are converted into frequency domain image by the Fourier transform. Gabor-filter can enhance those sub-blocks based on their main orientation in the frequency domain. Morphological operation Bot-Hat is used for extracting feature lines in different directions of the enhanced images. The test shows that this method can extract the palmprint feature lines information efficiently.Two feature extraction methods are proposed in the part. One is to convert the enhanced image from a spatial domain image into a frequency domain image and represent palmprint features using R-feature in the frequency domain. Another is to extract the palmprint feature lines using morphological operation Bot-Hat and represent palmprint features using the lines' geometrical feature. The experimental results shows that the two features can represent palmprint pattern classes effectively.A two-stage method of frequency domain features matching and spatial domain features matching is put forward. According to the R-feature, we can extract the classes that have similar features with the recognition-waiting sample. Then we start accurate matching using the geometrical feature.Experimental results show that the methods are effective. And the palmprint recognition system we have designed is feasible.