摘要
在掌纹识别中,角点检测是其中一个非常重要的环节,能否正确地检测出有用的角点对后续掌纹的识别影响重大。目前角点检测的方法虽然很多,但大多比较复杂,或检测出的角点数目太多,并且会伴随着伪角点,应用时需要人为去选择角点,对后期掌纹的识别带来很多不必要的麻烦。本文结合极值的理论,结合手掌具有的特点,提出一种新的角点检测的方法。该方法检测准确,无伪角点和偏移角点,在掌纹识别中,不需要人为去更正和选择角点,并使后续提取手掌中心点及手掌旋转的实现更加简便。
Comer detection is one of the very important parts of palm print recognition, which makes a great influence on the cor- ner detection. This paper provides a new method for comer detection combined with the extreme value theory and the characteris- tic of the palm. First, it uses the Sobel algorithm to split the palm image and get the binary image. And then it clusters the binary image, statistic gray value of each line of the palm image to locate the comer. Comparing to the classical methods, it makes a simple way to the center of the palm and the hand rotation.
出处
《计算机与现代化》
2012年第9期123-126,共4页
Computer and Modernization
关键词
角点检测
掌纹识别
HARRIS
聚类
comer detection
palm print recognition
Harris
clustering
