Finger Recognition by Feature Length

Department of Industrial Engineering & Technology Management, Da Yeh University, #168, University Road, Dacun, Chang-Hua, 51505, Taiwan

^* Corresponding author: cls2@mail.dyu.edu.tw

Abstract

A small circle cake in a finger image is used to extract the partial finger image. Subsequently, the major axis length of this extracted circle cake is computed. Next, more circle cakes are extracted and the major axis length of each individual circle cake is computed. These obtained major axis lengths are chained together circularly to generate recognition features. These features are used to recognize a finger. The chain-generated feature is compared to its adjacent features to locate the most match one. For every pixel in the sample image, the address of the match target in the unknown image is recorded. The trajectories from the sample image to unknown image can be similar and can be diverted significantly depending on the similarity of the two compared images. If the trajectories are similar, the match targets can hold together. Otherwise, the match targets are deviated to different places. The convergence and deviation situation can be used as an indictor to show whether the two compared images are coming from the identical source or not.