Research On Some Of The Technology Of Fingerprint Identification

Due to bargain price, miniaturization and simple operation of fingerprint identificationsystem, it is widely favored by the people and has become one main technology of the biometricidentification. Although the research of fingerprint identification has made some progress, butrecognition speed and recognition accuracy are still far from satisfied with the practicalapplication. And universal use haven’t been carried out. In this thesis, we focus our research onthe fingerprint image segmentation and normalization, the fingerprint image orientation fieldcomputation, and the fingerprint image enhancement.We have acquired the following findings:1) The traditional normalization algorithm can make the shallow ridges and the darkerridges in the fingerprint image to achieve a consistent level, we improve the normalization byusing block attributes to dynamically adjust the normalized parameters. The mean and varianceof sub-block,which acquired by improved the segmentation algorithm based on variance,can bedirectly used to the improved the normalization algorithm. And it avoids double counting oftime, achieve a perfect fusion at the same time.2) Researched with the main orientation calculation algorithm, the orientation based ongradient method which be smoothed have deviation with the original orientation. The orientationbased on mask method have high accuracy, but time consuming is long. We improve maskmethod to calculate the orientation of fingerprint utilizing characteristics of fingerprint ridge.Orientation of fingerprint which locate the upper left side and upper right side of the area has acertain range, so we simply calculate the orientation to remove the unreasonable direction andsave computing time consumption.3) Due to Gabor algorithm relies heavily on the accuracy of orientation field of fingerprintimage, we proposed PCA (Principal Component Analysis) algorithm for enhancement. Thealgorithm requires the directional decomposition of the fingerprint image, PCA filter can beacquied by training from the decomposition of fingerprint image. The simulation results imshowthe PCAenhancement algorithms can improve the accuracy of feature point extraction.4) Finally, make a summary of the full paper, and put forward some prospects of furtherresearch.