A Novel Fingerprint Bifurcation Extraction Algorithm Based on Neural Network Template Matching

Fingerprint feature detection is one of the key techniques of automatic fingerprint identification. There are some problems of rotation and shift in the present fingerprint feature detection. This paper has given an algorithm of bifurcation detection based on neural network template matching. Correlative matching is to calculate the correlative value between template and target image according to a certain criteria when the template moves on the target image. The center of best matching is considered as fingerprint feature. The criteria of scaling the matching value is regarded as optimization function. Thus, matching recognition is converted into function optimization. The designed matcher based on neural network template can keep rotation invariant when the target image has the rotation of a certain angle. Finally, bifurcation can be extracted by multi-input and single-output there-layer feed forward neural network. The experimental results based on FVCX, have shown that the algorithm has rotation and shift invariant.