| 2. | The thesis aims to solve the problems in the information acquisition and processing , feature extraction and classification in the hand back based identification . the main contents are as follows : ( 1 ) manufacturing a hand back acquisition device for hand back based identification , which can reduce the rotation and displacement of hand during the hand back images are captured , and build up a hand back image database which contains 600 images . ( 2 ) researching a novel 2dpca algorithm which extracts both horizontal and vertical is also proposed to choose and classify features
本论文旨在解决手背识别中的信息获取与处理,特征提取与分类等各种问题,主要内容包括: ( 1 )针对手背识别研制了一套手背样本采集设备,有效地降低了采集过程中手的位移和旋转对识别性能的影响,建立了一个包括600张图片的手背样本库; ( 2 )研究了一种新的2dpca特征提取算法,该方法可以有效提取水平和垂直方向上的图像特征,并对特征进行遴选和降维; ( 3 )在对chien的nfs的修正和改进基础之上,研究了一种新的nfs分类方法,该方法可以有效地去除chien的特征空间所包涵的冗余和不确定特征点,还解决了传统nfs特征空间的坐标原点问题; ( 4 )采用新的2dpca特征提取算法和nfs分类方法在手背样本库上进行实验。 |
