Research Of Palmprint Identification Algorithm Based On Gabor Wavelet Combined LBP

Palmprint recognition technology is an emerging biometric technology, which is one of the hot researches in recent years. Compared with other biometric identification methods, palmprint recognition has the following advantages:richer information, acquisition of palmprint image easily and requirement of low-quality, implementation in hardware easily and friendly user acceptance, fast recognition speed and easy integration with other biometric identification. Therefore, Palmprint recognition is an identification method of great potential for development.Palmprint recognition system and its development history and current research are studied and analyzed in detail in this thesis. In the preprocessing of palmprint image, a new extraction method based on key points for region of ROI area is proposed, which can avoid the deficiencies of traditional method based on the hand’s maximum inscribed circle. Firstly, the coordinates of the palm’s edge is perceived by utilization of the raster scanning method, and then the graph of palm edge is drawn in the Cartesian coordinate system. Afterwards, in order to determine the two key points between the thumb and index finger, the derivative of its Y-axis decomposition map is computed. Finally, coordinate system is reconstructed by use of these two points, and ROI sub-graph in the center of the palm is intercepted, which is normalized by a128×128size.In the stage of feature extraction, the extraction algorithm based on Gabor wavelet and LBP operator is studied separately, then a new algorithm based on improved Gabor wavelet combined with multi-scale LBP operator is proposed. This method uses the Fourier transform measurement method to determine the direction of the palm’s three mainlines, and utilizing them as the direction parameters of Gabor filters to get Gabor wavelet transform for palmprint; then a multi-scale rotation invariant LBP operator based on uniform pattern by changing the size of the scanning window to extract palm-print features is designed after Gabor transform. Gabor wavelet can describe multi-scale, multi-direction texture information of palmprint images, but be susceptible to light effects; meanwhile, LBP operator can represent the local neighborhood relation information of palm, which has the tributes of gray invariance and rotation invariance; however this method is easily affected by noise, and only to depict texture with a single scale usually. The new method proposed in this paper by combination of Gabor wavelet with LBP can utilize the advantages of them, which will achieve better results.In the stage of palmprint matching, GA and SVM are studied; meanwhile, the optimal parameters for the SVM is selected by using of GA. The experiments based on CASIA palmprint database show that97.5%accuracy recognition rate can be obtained by this paper’s algorithm.A large number of simulation experiments have been carried out in Matlab, and the results show the superiority of the proposed method. Afterwards, a palmprint identification system has been developed by using VC++platform, which can realize palmprint preprocessing, feature extraction and palmprint recognition.