Research And Realization Of Embedded Fingerprint Recognition Algorithrm

In the current fingerprint recognition system,.collect fingerprint image from sensor on the foreground and transmit the image to background server is extensively used to fingerprint recognition system.The traditional USB Key is required to enter PIN code to be authenticated in the client.The essential personal identified information will leave on the environment of PC.All of these traditional authentication methods have dangers of exposing personal identified infonnation in the network and personal valuable information will be easily hacked.The disclosure of above information will result in unnecessary loss of economic property.In contrast,USB Keys which are carried on fingerprint recognition process all personal essential information in itself.So the effective authentication information will not leave traces in PC and network.This will be better to protect the user's privacy and information security.Domestic security chip LC698S is a no MMU chip with SPARC architecture.This thesis mainly introduce.how to allocate memory space with customized memory pool,to solve memory fragmentation while program running for realization of Fingerprint verification through the way of space of RAM within LC698S.For this embedded platform,this thesis put forward a dynamic allocate against resolution algorithm and optimize for classic preprocessing algorithm and recognition with comparison of fingerprint feature point's topology.to fulfill requirements of FAR and FRR,at meanwhile reduce fingerprint recognition time.A new enhancement algorithm is put forward to reduce the resolution of the original image,change the distribution of image pixels and the total numbers,thereby saving the image storage space which can be carried out in the RAM space within the chip.All of this can accelerate the image processing speed and enhance the w hole fingerprint recognition's security.In this thesis,the false rejection rate and false acceptation rate of embedded fingerprint recognition is 0.151%and 0.026%,both of which fulfill the design requirements.The fingerprint authentication time is the time of extracting the feature information and the time of matching with fingerprint template,and the value is the sum of the two.It is emphasized that the fingerprint authentication time does not include the time that the sensor collects and transmits the fingerprint image.Because the algorithm proposed in this thesis satisfies the one-to-many fingerprint matching mode,the upper limit of the number of fingerprints stored is 20,so the feature information extracted by the input fingerprint image needs to be matched with the fingerprint template in the fingerprint template library.The matching time also includes the fingerprint template searching process.