Research Of The Embedded Fingerprint Recognition System

Fingerprint is very important in identity verification for its exclusivity and invariability. Automatic fingerprint verification system is used in many areas such as automated banking, national ID card,electronic commerce.The traditional fingerprint recognition technology based on PC got its restrictions of application and promotion in many civil field,because of its hardly mobility and high power consumption.So the embedded fingerprint recognition system with low power consumption and high performance will have bright future.In this thesis, I compare the digital signal processor TMS320VC5402, the fingerprint collect chip AF-S2 with some other products within their categories,analyze their advantages and disadvantages.Based on these two devices, I design the hardware of the embedded fingerprint identification system. We divide the system into two parts: the processing module designed on DSP platform and the fingerprint capture module designed on AF-S2 platform. We design and debug clock module, external memory module, fingerprint capture module and serial communication module. This hardware system is reasonable and feasible. The practical value of this system is high.Based on this hardware system, the fingerprint recognition algorithms are designed in my paper. We mainly study how to calculate the direction and frequency of fingerprint ridges, enhance the fingerprint image by Gaussian filter, and binary fingerprint by the method of dynamic threshold in fingerprint image's pre-processing. We can enhance fingerprint image and get good fingerprint image by fingerprint image's pre-processing in this paper. Then extract the detail points of fingerprint image by eight-field, on the analysis of the basis of pseudo feature points, study the method of removing pseudo feature points. Additional, we have developed an feature point-matching algorithm.The matching of point is transformed into matching of vectors,which are composed of two feature points. This algorithm obtains a better effect in feature extraction, pseudo feature points filtering and fingerprint matching. It can meet the need of the embedded system much better. By the experiments of 150 fingerprint in fingerprint database,we draw the conclusion that algorithm's rates of repulsive identification and error identification are low.