小波模极大值的手掌自动识别系统定位法  

Orientation Method Based on Modulus Maxima of Wavelet Transform Used in Palmpr int Identification

作　　者：盛丽[1] 吕英华[1] 孔俊[1] 

机构地区：[1]东北师范大学计算机学院,长春130024

出　　处：《吉林大学学报（信息科学版）》2006年第5期548-554,共7页Journal of Jilin University（Information Science Edition）

摘　　要：提出了一种有效、可靠的手掌识别定位方法。该方法对无定位的手掌自动识别具有较好的定位结果。研究过程中采用平板扫描仪进行图像采集。对手掌图像进行旋转、剪切处理,从而为同一个人的不同手掌提供相同的参照系;将手掌边界像素与质心的距离(边心距)作为描述手掌形状的特征,计算边心距的多尺度小波变换的模极大值,并由尺度a=25的小波模极大值的分布确定基准点所在区间;在每一区间上选取唯一的候选基准点,再对候选基准点进行筛选,最终定位手掌的基准点。该定位方法具有算法简单、准确、计算速度快的特点,应用本方法定位出的手掌基准点不仅可以为掌纹的匹配提供参照系,同时也可为结合手掌的其他特征如指纹、手形、手指上的纹理实现基于多特征识别提供必要的参照系。本方法对旋转和平移具有不敏感性,能够提高识别的准确率。通过M atlab仿真试验证明了这种方法的有效性和正确性。A reliable and effective orientation approach used in hand verification is presented in this paper. It has the advantages of accurate orientation for automated hand verification without a docking device. In our work, a flatbed scanner is adopted as the input device for capturing hand images. By rotating and segmenting hand, provide same reference system for a person＇s different hand images. And then take the distance between the pixels on the hand boundary and the centroid （apothem） as hand-shape feature, and compute multi-scale wavelet transform modulus maxima of the apothem. By analyzing the modulus maxima of scale a = 2^5, identify the section containing the datum marks. Then choose an exclusive candidate datum mark within each section. Finally, locate datum marks by filtering all candidate datum marks. This approach characterized by simple, accurate arithmetic and fast calculating speed. The datum marks oriented by this method can provide the reference frame not only for feature matching, but also for combination of fingerprint, texture of finger and hand-shape to realize identification based on multiple features. This approach has the insensitivity to translation and rotation, and it also can promote the rate of accuracy. The effectiveness and correctness of this method is proved by Matlab simulation.

关 键 词：手掌识别 基准点 小波模极大值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]