Research On Key Algorithms Of Swipe Fingerprint Identification System

The fingerprint swipe sensors have been widely applied to the portable automated fingerprint identification systems based on DSP and the embedded technology. The primary key problem of the swipe fingerprint identification system is real-time registering and mosaicking of the fingerprint sequence images collected by the swipe sensors, and to obtain a whole image of fingerprint for the sake of subsequent processing under the condition of limited resources.Another research object is to develop a set of suitable tailor-made algorithms for the swipe fingerprint identification system. Based on the theories and methods of the image processing and pattern recognition, the corresponding key algorithms have been developed for the fingerprint image preprocessing and fingerprint matching, and the performance of the system has been improved effectively. The main completed research works are listed below:1. The C5515fingerprint development kit of TI Company was utilized as an experimental platform for developing the swipe fingerprint identification system. The main components include TMS320C5515DSP processor and ATW310swipe sensor. Based on this experimental platform, the algorithm has been developed to collect fingerprint images by swipe sensors.2. A novel algorithm based on the method of phase correlation has been proposed for registering and mosaicking the swipe fingerprint sequence images.It is well known that the method of phase correlation based on the2D-Fourier transform property of image translation possesses better registration accuracy and robustness, it is required that the two adjacent images should largely overlap in contents. The real fingerprint sequence images don’t meet this requirement. Based on the continuity of finger movement, by means of the comprehensive algorithms of phase correlation and gray correlation, the problem of registering two adjacent images has been turned into the problem of registering two modules, and this difficulty has been overcome effectively.3. A series of key algorithms of the swipe fingerprint images preprocessing and matching have been developed and improved effectively.(1)The methods of the grayscale histogram equalization and the relaxation median filtering have been utilized to improve the quality of the preliminary fingerprint images.(2) The comprehensive method of combining gray variance with gradient orientation is utilized for the fingerprint image segmentation, and the corresponding algorithm has been developed to determine the segmentation threshold automatically.(3) The parameters of frequency and orientation have been determined reasonably, and the Gabor filter is used to enhance the fingerprint image effectively.(4)The algebraic geometry method based on matched segments in the central region was utilized to develop the algorithm of fingerprint matching.The above developed algorithms have been programmed and operated on MATLAB for inspection. Finally, the fingerprint matching algorithm was tested on the Fingerpass database for fingerprint cross matching setup by the biometrics research team at the Institute of Automation of the Chinese Academy of Sciences. The DET(Detection-Error Tradeoff) curve has been drawn based on the experimental results. The Equal-Error Rate (EER) is0.048, ZeroFMR is0.103, and the average matching time is0.28s. It has shown that the algorithms developed in this thesis can be applied to the AFIS based on DSP and swipe sensor. The research results can provide valuable practical reference algorithms and experimental data for the further development.