Research And Application On The Pre-processing Algorithms Of Low-quality Fingerprint

Fingerprint recognition technology is widely used in the field of identity recognition and identification with the development of modern science. The application of fingerprint recognition in the field of criminology is now developing from manual match to automatic identification. Because of the special input environment, fingerprints used in the field of criminology are always latent. Some of them have low contrast while others are incomplete with most parts of the ridge information losing. Currently, there are few researches on low-quality fingerprint images, and there seldom has a set of pre-processing algorithms that can improve the quality of this kind of fingerprints efficiently.The paper conducts some researches on the pre-processing algorithms based on the latent fingerprints with low contrast. Smoothing the noise before the common pre-processing step is recommended in this paper. In order to solve the problem of losing the edge information after smoothing, a nonlinear extrapolation filter is used to enhance the fingerprint images. This method recovers the losing edge information of fingerprint image. As to the segmentation part, a segmentation method combining the characteristics of statistic and the ROI region is used. The method solved the problem of over-segmentation of the foreground, and it has good effect upon the low-quality fingerprint images. By studying most pre-processing algorithms, a set of pre-processing algorithms that suitable for processing the low-quality fingerprint images is used. These algorithms include orientation extraction, estimation of ridge frequency, circular Gabor filter, binarization and thinning. To our knowledge, the burrs in the thinned image will lead to produce false characteristics. A method based on the rules is used to deal with the burrs after the analysis of the thinned images. The result shows that the method reduced the false characteristics efficiently, and benefited the matching step.The research is based on the development of constabulary automatic fingerprint identification system, so the paper is not only researching on pre-processing algorithms but also constructing the whole structure of the CAIFS, and embedding the algorithms mentioned above.The pre-processing algorithms are programmed under the VC++ platform, the results show that the pre-processing algorithms used in this paper can efficiently improve the quality of the latent fingerprints used by police, and give a good basis for the next matching step.