Design and Implementation of Palmprint Recognition System Based on Intel Atom
|Course||Computer System Architecture|
|Keywords||palmprint recognition system preprocess wavelet transform BP neural network|
In the era of rapid development of information, information security is growing more and more important, and it is a challenge for countries all over the world. Password, IC cards and other traditional security authentication method have been unable to meet people’s needs. Biometric security authentication technology comes at the right time, and it has become a major research and application direction. Among the existing biological characteristics, palmprint has been a hot spot, because it’s simple to collect, full of information, easy to user acceptance, and not easy to counterfeit, etc. Palmprint recognition has broad application prospects. Palmprint recognition technology is still at the stage of rapid development; therefore, palmprint recognition system has great significance in the practical application.This thesis not only introduces the development of the palmprint recognition in the past and the recent years at home and abroad, by reading a lot of relevant books and information, but also compares the mainstream palmprint recognition algorithms and implementation schemes. On the foundation of mastering key technologies, this thesis designs and implements a set of embedded palmprint identification system.This system runs on UP-Atom510embedded platform, and captures palm image by HD camera under semi-closed environment. Palmprint recognition software system is developed based on MFC dialog wizzard, using Visual Studio2008. Palm image process algorithm preprocesses the palm image, corrects the palm direction, locates a crucial area and intercepts a standard ROI (Region of Interest) area, based on principles of biological morphology. According to the natural multi-resolution characteristics of palmprint, the system uses Haar wavelet to transform the ROI area, and then subtracts the characters energy at different resolutions from the transformed image; then, system stores the enrollment information to Microsoft Access database. At last, this system trains a BP natural network as the classifier of the system, and uses it as the system identifier.The system has a friendly interface, and it is easy to operate. It has high real-time feature and robustness. Experimental results show that the system not only implements function of user enrollment and identification, but also reaches high recognition accuracy.