Research On Algorithm Of Automatic Fingerprint Verification System

With the development of automation and information, an accurate and high effective personal identification is more important. Today,the traditional methods of identification become more limited and more faulted,whereas Biometrics becomes the new development trend of the personal identification owing to its particular advantage. Automatic fingerprint identification system, a technology of fingerprint identification using the computer, is the hotspot of Biometrics. This thesis is mainly about the study and design of the Automatic fingerprint verification system applying for one to one match, including the key algorithm.The object of this thesis is to propose an algorithm of fingerprint verification for real time processing, enabling shorten the verification time while the accuracy does not reduce. The algorithm behaves as a system.This thesis studied the current about the algorithms of fingerprint verification, summing-up two dominant fingerprint verification methods: minutiae-based and filterbank-based, and studied the key in design concretely. Thereby, the scheme in the thesis, verifying fingerprints based on the wavelet transform and Gabor filter method, has been proposed.The fingerprints were viewed as oriented texture patterns. Their structure was different in whole figure, but showed lowpass characteristics along the ridges or valleys and bandpass characteristics along a direction normal to local ridges orientation in a local neighborhood. Therefore, a bandpass filter that is tuned to the corresponding characteristics can efficiently enhance the structures of ridges and valleys. On the other hand, Gabor filters have both frequency-selective and orientation-selective properties and have optimal joint resolution in both spatial and frequency domain. Concerned studies showed it is appropriate to use Gabor filters as bandpass filters to enhance fingerprint figure and simply the feature representation and match of fingerprints. Therefore, a filter-bank composed of 8 Gabor filters which have the same core frequency and the different direction was used in this thesis, which was tuned to the corresponding characteristics of fingerprints. Experiment results show it is able to enhance the structure of ridges and valleys by using the filter-bank filter the fingerprints. According the results, the thesis analyzed the influence of parameters (and)associated with space envelope. According to open literatures, about 99% of the total compute time for verification is taken by the convolution of the input image with 8 Gabor filters. Therefore, it is the key to improve the verification speed by reducing the calculation. Based on the idea, this thesis proposed to add wavelet transform before using Gabor filters to compress the image data, accordingly to decrease the calculation. Because wavelet transform carried out filtering in essence, the layer number or decomposition and selection about wavelet function have both influence on the accuracy and speed of the verification system. The thesis verified the fingerprint based on its structure feature, so the compressed image data can not lose the feature before them were filtered by Gabor filters. The thesis analyzed theoretically it meet the requirement that one layer orthogonal wavelet decomposition was applied to the subject. The approximation of the original fingerprint image was selected to replace the original, and filtering it can decrease the verification time. The experiment result showed selecting wavelet db9 can attain more excellent speed and accuracy.The thesis also studied one of the key steps in fingerprint identification: the orientation fields estimation. The improved method based on the gradient was selected to estimate the orientation fields in the thesis. By the experiment using the method before improved and improved respectively on MATLAB6.1, it showed the selected method can accurately estimate the orientation in the fingerprints.The thesis realized the algorithms directly using Gabor filter and using Gabor filter combined wavelet transform, and...