Palmprint Feature Extraction And Recognition Method Research Under The Rigid Deformation Condition

The systems and information security in the modern society is increasing more and more important.So are the occasions to identify the identification. Biometrics becomes more widely used for its unique stability, uniqueness, convenience. Palmprint recognition is a focus research. It becomes a very potential identification method in recent years. This article first elaborates the history of palmprint recognition, analysis of its research status and some of its advantages compared to other biometric. Then study the hand deflect problem when the palmprint image acquisition process without using any fixtures.Design a hand normalized treatment to correct rigid deformation palmprint images. Established different deformed hand images library. Palmprint recognition method is used to feature different degrees deformation palmprint images and the corrected images.In the preprocessing module, deformation palmprint images are normalized corrected first. Resist palm image distortion caused by errors on the recognition by adjusting the aspect ratio of the different hands to the same size, positioning the contour points on the corrected image to extract the region of interest again. This paper attempts to locate by three methods. All the methods were compared in the library experiment.In the feature extraction module, studying 2DGabor,2DPCA and two line features methods. Dynamic threshold binarization method and the classical edge detection method are used to extract the main plamprint line.Change the palmprint line binary image into a single pixel line and more smooth image eventually by enhancing, breakpoint connecting, refinementing, deburring.In the classification module, calculating the Euclidean distance between test images and training images when using 2DGabor,2DPCA. Determining the image belongs to which class by observing euclidean distance corresponds to the minimum number.The method to classify the line feature image is to rotate the input image in 4 directions, obtaining template 1 by operationing and smoothing after or swelling.Obtaining template 2 by shifting the image in 8 directions.Using a template to match the input image, getting a relative better recognition results.