中文摘要：

针对现存的签名鉴伪方法有效性低、鲁棒性差的问题,提出一种基于图像特征自适应融合识别的签名鉴伪方法。该方法首先对签名图像进行预处理并提取签名常规性特征和签名鲁棒性特征,然后对签名字符书写形态、单方向笔画重量分布、字符像素点分布、图像字符倾斜角度等图像信息进行检测,通过分析检测结果自适应地改变各特征的权重系数进行特征融合。利用待鉴伪签名融合特征与数据库融合特征的向量夹角相似性度量结果对鉴伪样本做出判断。实验结果表明该方法在签名鉴伪中具有良好的有效性、鲁棒性。

英文摘要：

In view of the low validity and poor robustness of the existing authenticity identification method for signature,this paper proposes a method to indentify the authenticity of signature,based on self-adaptive fusion of image features. First,a signature image is pretreated,and normal features and robustness of the image are taken. Then,the writing form of characters,weight distribution of unidirectional strokes,pixel dot distribution of characters,inclination angle of image characters and other image information are inspected. By analyzing the inspection results,this method automatically changes the weight coefficient of each feature and fuses the features. Then this method judges the measurement results of the vector angle similarity between the fused features of signature to be identified and fused features of the data base. Result of the experiment shows this method has good validity and robustness in identifying the authenticity of signatures.