The Research Of Whole Palm Vein Recognition Algotithm

As a kind of computer security technology，biometric identiifcation technology hasrapidly developed in recent years. Extracting and a seires of processing of humaninformation such as fingerpirnt, iirs，face，and other physical information are used torecognize and identiifcate. Compared with the traditional password, key, etc.，biometricidentiifcation technology has advantage of not easily being changed or lost，and highersafety. Vein recognition is a kind of new biometric identiifcation technology，which couldbe separately as a kind of feature recognition, and can also associate with hand shape,palmprint traits to form hand multimodal fusion recognition.This paper&st birelfy introduces the whole hand vein recognition system overalldesign; and then it introduces the image acquisition device such as camera, the wavelengthand intensity of light and according to the environment choosed, establishes threeexpeirmental image databases. And an extraction method for finger valley points undernatural state is proposed，which can avoid image binairzation and contour extractionoperation, improving the running speed of this algorithm. A self-built image database(SUT，which consists of420images) is used to test this algorithm, and the correct extraction rateis97.14%, proving it feasible and effective. And this method can avoid the impact offingers’ opening degree, proving it superior.The next chapter descripts detailed image preprocessing in vairous steps: extract threevalley points, extract other feature points, set up coordinate system according to extractfeature points, normalize images(including size normalization, direction normalization,coordinate origin normalization), extract palm area，normalize the gray-level of palm area,and then denoise. Then segment and process the vein structure, analysis the segmentalresults. The last step uses several classic methods to extract feature of whole palm veinarea, and each method is described in detail, and ifnally chooses a feature extractionmethod based on gray level: the statistical characteirstics of the method，and uses the Euclidean distance minimum identiifcation to match. The recognition rate of wholealgorithm in embedded systems reaches about70%.