| Automated fingerprint identification has become an exciting area of interest for researchers in the past several years. Since individual fingerprints are widely accepted by scientists and courts, commercial systems for fingerprint recognition have been developed accordingly. However, due to the limited technology in fingerprinting techniques and the computationally intensive nature of recognition, the lack of a reliable identification system is still an existing problem.; In this study, a cost-effective automated fingerprint identification system is proposed and tested. The model uses computer-scanned fingerprint images as inputs, removes the noise from the images using spatial and frequency-domain techniques, extracts the ridge orientation of each fingerprint, classifies them, and then among the categories, matches them with the correct fingerprints.; A hybrid two stage neural network architecture, consisting of hierarchial neural networks combined with a rule-based syntactic network, is used for classifying and recognizing the cleaned fingerprints. In the first stage the neural network classifies the fingerprints according to their general characteristics. The identification stage of the model uses the minutia information.; The results are encouraging for industry applications. The system provides unique identification for small databases. The future developments for such a system can be use in companies or facilities where security is a priority. |
