Unsupervised Fingerprint Classification with Directional Flow Filtering

In this study, an unsupervised neural network model is proposed for fingerprint classification. The proposed model uses directional flow or local ridge orientation (LRO) information and the relative locations of the fingerprint singular points as features for the neural network. The LRO is obtained by storing the directional data in a neighborhood from the raw fingerprint and this data is filtered to prevent the flow errors due to noisy sections of the scanned fingerprint. From the resulting LRO data, the singular points are extracted and the information is used as the input to the SimNet unsupervised neural network model and the fingerprint is classified accordingly. For identification, the raw image is passed through a band-pass filter and the minutiae are extracted from the processed and thinned fingerprint image. The fingerprint identification is performed within the obtained class by comparing the minutiae of the input fingerprint to the other existing prints in that class. The number of classes can be adjusted by the number of the singular points and the distance between these points. The results show that the model can generate a hierarchical classification model that can reduce the search space for identification which could be used in small to mid size fingerprint database applications.