Fingerprint Identification Based On Wavelet Transform

With the social development and progress, identification is widely used in various fields, how to accurately and efficiently take its authentication has become a hot research topic. So the technology of biometric fingerprint recognition has been taken the attention by a growing number of researchers. In practice, due to the device or the poor quality of fingerprint images, a higher demand on the algorithm in fingerprint identification system is proposed.This dissertation is based on the wavelet theory and pattern recognition, using the wavelet features of the fingerprint image to enhance and extract the image feature and combining the method of probabilistic neural network to match the fingerprint characteristics to complete fingerprint recognition.The study contains three parts:fingerprint preprocessing, fingerprint feature extraction and fingerprint feature matching. The major work of fingerprint preprocessing is image enhancements, normalization and equalization, image enhancement, image binarization and refinement to facilitate the feature extraction and matching. Fingerprint enhancement based on wavelet transform of adaptive threshold method. Extracting feature vectors is the main work of fingerprint feature extraction, the main job of which is selecting the fingerprint image to identify areas with the four-layer wavelet transform to extract a pair of12-dimensional feature vector. The fingerprint feature matching is to match the extracted feature vector and fingerprint databases, the main work of which is to match with the application of probabilistic neural network, to search the optimal value of the smoothing factor based on the probability density function estimation and Bayesian decision theory.In the end, this dissertation makes numerous experimental analyses on the algorithm for fingerprint identification based on wavelet group of Dubieties Series, at the same time, analyses factor of the algorithm performance, for example, the recognition rate、 reject rate、misclassification rate and recognition time and so on.