Study On Invariant Feature Based Fingerprint Recognition Algorithm

Fingerprint recognition is an important and challenging area in digital imageprocessing. With the increase of authentication and person’s concern, fingerprintrecognition is still attracting more scholars to research. Fingerprint recognitionalgorithm is sensitive to fingerprint translation and rotation, due to the noise andnonlinear deformation. For the translational and rotational invariance of thefingerprint recognition algorithm, fingerprint recognition based on the invariance isstill a very important issue.The paper focuses on how the recognition algorithm realizes the invariability tothe fingerprint translation and rotation, and we pay great attention to the matchingalgorithm based on filtering features and minute features, including two aspects:1The recognition algorithm based on the eight directions Gabor filters. Weenhance the low quality image based on the Gabor filters in order to improve theimage quality. The center point location is based on the complex filter, and themethod has good effect on the low quality image, soling the fingerprint translationinvariance. In the feature matching, by circulating the double finger code, andcompared with the finger code to be matched more times, we solve the fingerprintrotation invariance effectively.2The recognition algorithm based on the local structure. The key point of solvingthe fingerprint translation and rotation invariance lies in the accuracy of thetranslation and rotation calibration parameters. In order to get the real and effectiveminutiae, we firstly use short time Fourier transform to enhance the low qualityfingerprint, and use the ridge tracing method combined with the character of thefalse feature point to do post–processing,in order to get the real and effective point.We build the eigenvectors of the local neighborhood, build the matching scoresystem according to the deformation impact on the feature, and overcome the impactof the translation and rotation. We use the bounding box of variable size to solve thefingerprint local nonlinear deformation problem.In the paper, two algorithms are tested and simulated on the plate form of MATLAB7.0.The results show that algorithm can realize the invariability to thefingerprint translation and rotation well. The paper has achieved our predicted goaland meets the research’s requirements.