Fingerprint Recognition Based On Neural Networks

With the development of the computer and information technologies, identity identification technology by computer makes quiet great progress. Because of the invariability and uniqueness of the fingerprints, the technology of fingerprint verification and identification has become one of the widest used identity identification technologies. It relates to a lot of fields, including pattern recognition, digital image processing, digital signal processing, artificial intelligence and so on. It has great worth in criminal identification, network safety, information facility safety etc, and has not only theoretical significance but also practical value.Now, Automated Fingerprint Identification Systems (AFIS) is still in the stage of experiment. What makes it far from mature is often ascribed to its demand on large numbers of fingerprint-matching processes, and the considerable influence of images' noise and skins' elasticity during the procedure of automated fingerprint identification. These limitations result in a low recognition rate and a slow fingerprint-matching speed. In this paper, we try to combine the neural network with pattern recognition and find several new methods in fingerprint automatic identification as the practicable solution to the trouble.In general, we mainly research on fingerprint preprocessing, feature extraction and neural-network-based fingerprint identification in this paper.In the procedure of preprocessing, we research on the operations of enhancement, segment, binarization, thinning as well as normalization, and then achieve a program for its application. We design a normalization operation as the additional step to the fingerprint preprocessing operation. The step of fingerprint normalization not only normalizes the size of fingerprint image, facilitates the test on AFIS, but also reduces the AFIS running time and fastens AFIS recognition speed. Additionally, this paper does a pilot study on fingerprint thinning algorithm which is based on neural network.In the procedure of fingerprint feature extraction, classifying feature and identifying feature have been extracted. For fingerprints classification, we use fingerprint's direction feature to classify a fingerprint and get a desirableclassification result. For fingerprints identification, we present a fingerprint's local minute feature to identify a fingerprint based on neural network and a good performance has been achieved, this minute feature give full information of fingerprint's details.In the procedure of fingerprint identification, we divide the process of fingerprint identification into two stages that use different neural network models to classify and identify fingerprints. First, the BP neural network has been used to classify the fingerprints into four categories, and then based on their category, some fingerprint's minute features have been used for the fingerprint identification with the LVQ neural network. Finally, some experiments have been made for our AFIS and satisfactory recognition results have been achieved.