Research On Filterbank-Based Fingerprint Identification

This thesis focuses on filterbank-based automatic fingerprint identification technology. We discuss and experiment on locating reference point, designing Gabor filters, filtering and constructing, extracting features and matching. A two-times locating method based on the gray contrast of surroundings is presented in locating central reference point. This method has better adaptability and robustness, and can basically meet for the accuracy of automatic locating center. Fourier spectral analysis is improved to obtain a more accurate ridge frequency in computing ridge frequency. The average ridge width is used in designing Gabor filters, so the filters have certain adjustability. The rotating invariance of fingerprint images is solved by the rotating of feature models in feature matching.Fingerprint image segmentation is discussed. We design a dynamic threshold based on gray characteristic which performs well. We also proposed a dynamic threshold segment arithmetic based on energy which is simple, good and adaptive for different quality.Fingerprint classification is also researched on. We select a method based on direction changing of central area after analyzing fingerprint images and relative arithmetic etc, and change the size of cetral area. Our experiment shows it is simple and fast.We develop a filerbank-based automated fingerprint identification system. It works well with good recognition rate, stability and robustness, and can meet for special civic application.