The Palmprint Image Acquisition And Processing System Based On DSP

With the rapid development of informatization, digitization and anonymity ofpersonal identity is becoming a trendy in this research field. How to determine theidentity and secure personal information is a significant topic. Some previousresearchers proposed physiological features identification methods and obtainedoptimistic results. The identification methods satisfy special requirements based onthe unique biometric feature difference which can be easy carried but difficult to bestolen. Among them palmprint recognition draws attentions of some researchers for itsrich harvest of information and the stability of character.The theoretical palmprint recognition methods are introduced, and thecorresponding identification products also are presented in several public places.However, these technology and products are always designed on PC, with thedisadvantage of huge volume and high costs. It is not beneficial to the development ofpopularization and commercialization. Considering about the current developmenttrend of the palmprint recognition method, the embedded recognition system has abroad commercial prospect. For example, the embedded palmprint recognition systemplans designed based on ARM and DSP are always adopted for hand-shape imageacquisition, processing and recognition.Focusing on this research topic a DSP based palmprint recognition system isproposed in this thesis. Emphasizing on the acquisition and processing algorithmdesign of palmprint image on DSP, the main research contents are as follows:(1)Combining TMS320DM642and TVP5150A, the data of palmprint imagefrom the analog signal offered by CCD is decoded with VP5150A. And then a DM642is used to read the data and progress the following treatments to accomplish theidentity recognition.(2) Palmprint image acquisition based on DSP/BIOSDSP/BIOS is an embedded real-time operating system which is speciallydesigned by TI Company as a real-time operating system kernel for the DSP chip of TI. It can facilitate the multi-task programs design for users. On the basis of readingsome materials about the DSP/BIOS, the class/mini driver model and API functionfrom the chip driver development kit (DDK) are used in the acquisition programdesign to complete the palmprint image acquisition.(3) Palmprint image processing based on DSP/BIOSSince there are differences on the format of data and the storage method betweenthe DSP chip and PC, the processing speed of DSP is slower than PC. So for theprevious algorithm some adjustments and modifications must be done to for thereal-time operation of DSP system. So some conventional and lower complexmethods are always used in palmprint image processing. In this thesis a3*3fastmedian filtering improvement algorithm is proposed. A conventional and lowercomplex method for threshold segmentation and contour extraction is adopted. Incorner detection stage a relative distance-based detection method with fewercomputation and higher accuracy is proposed. In the process of feature extraction acombination algorithm of wavelet,2DPCA and KPCA is used to obtain highrecognition rate. Based on these methods the processing effect and the real-timequality can be taken into account in the palmprint image processing.In the end, the experimental results are summarized. The real-time quality andthe CPU processing ability of the system are analyzed as a reliability proof of thepalmprint image acquisition and processing system designed in this thesis.