【作者】 李媛；

【导师】 卡米力·毛依丁；

【作者基本信息】 新疆大学 ， 计算机应用技术， 2008， 硕士

【摘要】 笔迹鉴别是依据手写笔迹对书写人身份进行判别的一门技术。随着计算机技术的发展,笔迹鉴别的重要性在当今的社会越来越明显。在工作中,各国政府之间的协议和备忘录的签订,官方文件,以及法令和法规的颁布,银行支票的签署汇兑,都需要当事人的亲笔签名才有效。在实际生活中,社会交往信函,则不管是手写还是打印,最终也是作者本人的亲笔签名。用计算机实现笔迹鉴别可以减轻文检人员的工作压力。笔迹鉴别已成为计算机视觉和模式识别领域研究的热点问题。本文首先分析了笔迹鉴别的发展历史及应用背景,并介绍了目前的技术发展状况。以笔迹书写人身份鉴别问题为背景,研究基于改进型的多通道Gabor小波变换的笔迹鉴别问题,建立了基于该算法的笔迹身份鉴别系统。本文工作涉及:笔迹图像的预处理,笔迹图像的纹理特征提取,笔迹图像的分类匹配,以及基于改进型的多通道Gabor小波变换的笔迹鉴别系统的构造和实现。通过实验测试,系统准确率可以达到70%,基本完成了系统设计目标。系统编程平台为VS.net 2005。本文取得的主要研究成果如下:(1)在改进型Gabor小波变换算法的基础上进行实现方案上的改进,提出了适合维吾尔笔迹特征提取的纹理分析算法。该算法与笔迹所写内容无关,可记录下每一个通道的书写人笔迹风格的特征向量――均值和方差,它们记录下了每个通道笔迹图像纹理特征的重要信息。(2)采用最近邻距离和对笔迹样本进行分类匹配,将训练后的32通道64维特征向量以纯文本格式保存于相应的笔迹库中。其优点为数据存储量小、便于管理,辨别速度快。更多还原

【Abstract】 The handwriting identification is a kind of technology used to identify the figure of a person’s. With the development computer technology, the importance of handwriting identification is becoming clearer and clearer in today’s society. The agreements and memorandum between the governments signed, the official documents, laws and regulations, bank checks signed and exchanged are all effective by the autographed from both parties. In real life, the social despondence letters must be signed by the author himself, no matter it was written or printed. Using the computer to identify the autograph can reduce the pressure of the staffs’. It also becomes a key problem in the computer vision methodology and pattern recognition.This paper firstly analyzed the history of development and the application background of handwriting identification, and introduced the current condition of it. This paper considered the reorganization of writer’s identity as the background, the research based on the problem of improving the multi-channel Gabor wavelet transform, handwriting identification based on improve the multi-channel Gabor wavelet transform , established the writing identification system of its own algorithm. This study involved handwriting image preprocessing, texture feature extraction of the handwriting image, handwriting image classification and match, and the construction and achievement of the improved handwriting identification systems with the multi-channel Gabor wavelet. The experiment showed the rate of correctness is 70%, which achieved the goal of the system design. The system is accomplished in VS.net 2005.The results of this paper are as follows:(1) On the improvement of the Gabor wavelet transform, it presented the suitable texture analysis feature extraction algorithms in the Xinjiang Uygur handwriting. This arithmetic is a text independent method, which can record the mean and standard deviation of each channel. These are the important information of the textural characters.(2)This paper also designed the classification of the handwriting image, using nearest neighbor classifier to fulfill the identification task. Its advantages are small memory of data storage, manageable and fast identification.更多还原