The Fingerprint Image Preprocessing And Feature Extraction Study

Rising in the 1960's, the fingerprint verification technologies (FVT) identified aperson based on the intrinsic physiological information of fingerprints is becoming amain biometric identification technology. Because of the uniqueness andinvariability of the fingerprint feature and the feasibility of fingerprint verificationsystem, FAT have been the focus of the biometric identification technology recently.FVT include three major steps: preprocessing, features extraction and featuresmatching. Some research of the key algorithms of preprocessing and featureextraction and its implementation been introduced in this paper, the main contentsconclude:1.By means of the linear discriminant analysis (LDA), the segmentationalgorithm to separate the foreground and background of fingerprints is proposed sothat the feature extraction can be focused on the foreground of fingerprint, as a resultthe time for post-processing could be reduced.2. In order to fill the holes of the ridges and eliminate the small black blocks ofthe valleys, the method to reduce pseudo features is proposed in this paper.3.To the slow speed of fingerprint thinning, the method of twice steps thinningis proposed. The first step is finding the approximate position of the features in theridges of fingerprint, the second step is exactly localizing the position of featureswhich had been found in the first step. Through the method of twice steps thinning,the time for finding features can be reduced.4.The frame of fingerprint verification hardware system is built in this paper,and the filter algorithm which taken up a lot of time of processing and featuresextraction is implemented by FPGA. Usually, the filter algorithm was completed byDSP in the automatic fingerprint verification hardware system, but it is proved to betime consumed, so in order to cut down the time of system, the filter algorithmimplementation based on FPGA which compensate the deficiency brought bysoftware compute is attempted in the paper.The simulations of the above algorithms are preformed in the computer andprove the algorithms'effectiveness.