Design Of Non-contact Multi-spectral Handing Imaging Software System Based On DM642

With the improving of the domestic productivity, biometrics has gradually entered people’s life. This paper designes software system of a non-contact multispectral hand recognition device.The device can put the collected three-modal feature of the palm-vein, palm-print and palm-shaped together.This method overcomes not only the health problems caused by the touch of hand recognition device,but also the problem of poor single modal identification.The device also has the necessary for efficient safety identification system.In this paper,an embedded DM642-based software system was designed.The purpose is to achieve hand biometrics. The hardware structure of hand image device was given.The software system contains:collection,register and recognition,which was mainly concerned.Firstly,details on register configration and address space allocation were discussed.Then registration information was transferred from PC to DSP by Socket communication.Lastly,feature points were extracted to complete database creation.And registrant information was displayed on a LCD.After the experiment,the simulation results were recorded to find the conlusion.This design can achieve the real-time collection and information recognition.It uses the dual-image sensor(OV7740) to collect the hand-information, uses DM642 to complete the image processing. The data transmits from the VP port to the encoder chip, converts the signal D/A with coded chips, outputs analog signals into the LCD screen on display. The collected data comes through the hand image Ethernet media access controller(EMAC), process the data by the physical layer chip(PHY), sends to RJ45, completes the registration and other related work on the PC, and then the database is established.Finally,program for recognizing was burned into the FLASH of DM642 to achieve the off-line working.The system has achieved real-time display and data transmission. By embedding the correspondingly fusion-processing and recognized algorithms, it can be completed on the user’s identity and build the ideal platform.