Algebra And Geometric Features A Combination Of Fast Fingerprint Identification

Although research of fingerprint identification technology has made great progresses, it remains a hot topic in such fields as image processing and pattern recognition. Researchers focus on how to increase identification speed and accuracy.Good results of fingerprint preprocessing are the prerequisite for high accuracy and speed of identification. In light of highly dependence among steps of fingerprint preprocessing, we analyzed the whole process of fingerprint preprocessing, proposed new approaches for image preliminary filtering and the design of orientation filter, and got satisfactory results of preprocessing on the fingerprint containing gauss white noise. Because the high resolution on high frequency of wavelet package meets the feature of fingerprint containing abundant high frequency, we filtered the fingerprint preliminarily by means of wavelet package. While designing orientation filter, we introduced the square-circle transform to rotate filter, thus we reduced computing complexity and avoided imported errors.Traditional matching methods, only based on the geometric feature of fingerprints, are lacking in matching speed. To increase speed and keep higher accuracy simultaneously, we modified traditional methods by proposing a grading match method based on both the algebraic feature and the geometric feature of fingerprints. First we extracted the algebraic feature of fingerprints through the wavelet transform, then excluded most fingerprints that can't match the source fingerprint in rough scales, finally processed further matching based on the geometric feature on possible fingerprints. As a result, amount computing is greatly reduced, and speed is increased. In our experiment, matching time is reduced by 47.5% while higher accuracy is kept. The algorithm is expected to be a practical and effective fingerprint identification technology.