With development of technology, the accurate personal identification becomes more and more important and necessary. The use of biometrics in personal identification highly improves the performance, security and convenience of the identification system. As the oldest mode of biometrics, fingerprint identification system is the most prevalent in use today. This thesis focuses on algorithm of feature extraction related to Automatic Fingerprint Identification Systems (AFIS), and the main contributions are as follows: (1) A comprehensive preview is presented on fingerprint image preprocessing on the state of the core algorithm related to AFIS. (2) We present a novel fingerprint classification algorithm that is based directional fields. Then we extract features that we define from fingerprint images. We also use k-mean and 3 nearest neighbor to classify features and distinguish the fingerprint. This algorithm improves accuracy more highly than the algorithm which uses the singular points. (3) A new fingerprint image segmentation algorithm is proposed. We define two features: contrast and main energy ration in the special domain and frequency domain respectively. Then we use RBF neural network to perform training, classification and segmentation. Because by the use of this new algorithm of segmentation we can reduce the number of false minutiae extracted in the blurred area and improve the accuracy of feature extraction and the whole performance of AFIS. (4) We present a new algorithm of fingerprint enhancement. Because there are many differences in the normal ridge area and the singular point area, we must classify these two areas, design filters and enhance them respectively. Moreover, because the method of ridge width estimation is used, we get the better result of enhancement. (5) We introduce an embedded fingerprint identification system and the application of fingerprint image preprocessing in fingerprint lock recognition module.
修改评论