Biological studies have shown that fingerprints' ridge structure never changes throughout the life of an individual after they are fully formed at about 7 months of fetus development except due to accidents such as bruises and cuts on the fingertips. This characteristic makes fingerprints very attractive for biometric authentication. Automatic fingerprint identification system (AFIS) includes the acquisition and storage of fingerprint images, the representation and feature extraction, fingerprint classification and indexing, fingerprint matching and other modules. Aiming at some key technologies in each module for the large capacity of AFIS, we established optimization models, designed fast and accurate algorithms. Experimental results showed that our algorithms are robust and effective. AFIS embedded in our core algorithms has been applied to criminal investigation areas in some provinces and cities in China.