Palmprint Verification Based on Extracted Features Using SURF Algorithm and FLANN Classifier

Biometrics science is one of identification methods in which it is tried to determine identity of human based on his behavioral or physiological characteristics. Among different biometrics methods, palmprint is one of the most reliable ways to identify people because characteristics are extracted from large area of palmprint and don’t change during lifetime. Different techniques consider the challenge of increasing accuracy and reducing time. In this research, SURF algorithm has been investigated to identify identity which aims to investigate performance time of algorithm and to assess accuracy of its performance. In methodology of this research first, High Boost Sharpening pre-processing has been conducted on images, then feature extraction and description of key points have been done using SURF[1] algorithm and descriptions are saved in a matrix in classifier. Finally, key points of requested image are extracted and theses points are described, then their distance is calculated from each saved descriptions in previous step and correlated image of those descriptions which have the least distance gets chosen. FLANN[2] has been used for classification and PolyU palmprint database is used to assess algorithm capability. Experimental results on PolyU palmprint database show that accuracy rate of SURF algorithm gained 98%.