Skip to Main Content
A novel approach which use single knuckle-print image only to implement personal identification is presented in this paper. Unlike most previous work, there is no need to collect a large amount of images to train the classifier. The identification process can be divided into the following stages: extracting the feature of the knuckle-print and matching the line feature. For the texture characteristic of the knuckle-print, we use a self-defined convolution template as the gradient operator to carry out the edge detection and extract itpsilas line feature. Moreover, In order to solve the dislocation of the images between the matching images, a new method for line feature matching is proposed, which improved the correct identification rate at a large extent. The approach was tested on a database of 98 people (1, 579 Knuckle-print images). The experimental results showed the effectiveness of the method is obviously.
Information Processing, 2009. APCIP 2009. Asia-Pacific Conference on (Volume:2 )
Date of Conference: 18-19 July 2009