Research On Computation Methods For Fingerprint Orientation Field

The auto fingerprint identification system (AFIS) has been widely used in themodern society, especially in some security fields which need identificationauthentication. Compared with other identification authentication methods, thefingerprint recognition system has a great advance in popularity, stability, uniqueness,and it is easier to collect and more acceptable. Although the technology of fingerprintrecognition has become more and more mature, there are still many problems need toimprove, especially the recognition of low quality fingerprint and the recognition oftouchless fingerprint which is rising in recent years. The computation of the orientationfield also plays a crucial role in the whole fingerprint recognition system. The correctorientation filed is the basis of collecting and matching the characters of fingerprintimages correctly. The president methods of computing the orientation field can correctlycompute the orientation file of high quality fingerprint images, but these methods can beaffected by the noise, so they can’t compute the orientation of low quality fingerprintimages correctly. And the problem of computing the orientation field of the touchlessfingerprint images can’t be solved by traditional methods, too. In view of this situation,we raise two new methods to compute the orientation field of low quality fingerprintimages and touchless fingerprint images respectively. It can be certificated that thecomputation methods raised in this paper can compute the orientation field of touchlessfingerprint images and low quality fingerprint images, especially those with crease inthem. The major contributions and innovation are as follows:(1)A method based on morphological operation to regulate the orientation field isproposed. This method targets the problem of orientation field computation in lowquality fingerprint images. On the basis of traditional ridge orientation computingmethod, the method we proposed can revise the erroneous ridge orientation byimplementing morphological operation. The experience shows that this method canestimate the fingerprint orientation field correctly and robustly for low quality images,especially for those with creases.(2)A method based on auto adapting gradient operator to regulate the orientationfield is proposed. This method targets the problem of touchless fingerprint imageswhich have different ridge distance in different areas, implementing different gradientoperator in different area to regulate the orientation field on the basis of statistic modelof ridge distance for touchless fingerprint images. The experience shows that thismethod can solve the problem caused by ridge distance difference and regulate theorientation field correctly.(3)A statistic model of ridge distance for touchless fingerprint images is proposed.We pick a large number of fingerprint images for experience, extract the effective area, compute the ridge distance in the effective area and count the ridge distance ofexperience images, getting the relationship between the ridge distance and the ridgelocation, establishing the statistic model of ridge for touchless fingerprint images. Theexperience shows that this model can instruct the selection of gradient operator in thecomputation of orientation field based on gradient.