| Summary: | 碩士 === 國立中央大學 === 資訊工程學系 === 105 === Recently, biometric verification attracts tremendous attention and plays an important role in personal authentication. There are many inherent biometrics that have been widely used in personal authentication systems such as face, fingerprints and iris, etc. This thesis focuses on palmprint verification. Most of the traditional palmprint verification methods acquire palmprint images using contact acquisition devices, which is inconvenience in use and hinders the practicality in application. In order to improve these shortcomings, we propose a contactless palmprint identification method by acquiring palmprint images through contactless acquisition devices.
In this thesis, the inputted image is first pre-processed to extract the region of interest (ROI), and then scale invariant feature transform (SIFT) is employed to extract the features of the image. To solve the non-linear deformation in the palmprint image, we propose an automatic partitioning method to divide the palmprint image into multiple regions before performing the matching of feature points. The feature points of each region will thereby be matched separately. Finally, the matched points are refined by employing random sample consensus (RANSAC) and optimization within the area to remove those matched points which fail to satisfy the topological relationships. The number of final matched SIFT points is then taken as the score for decision. Experimental results demonstrate that our proposed method is effective and robust in contactless palmprint verification.
|
|---|
