用于离线英文签字识别的特征点匹配

Feature points matching for off-line signature verification

本文提出一种用于英文签字识别的特征点匹配算法。该算法通过比较各特征点的局部特性及特征点之间的相对关系，计算各特征点间的相似性从而得到特征点匹配关系。特征点集通过沿笔划骨架的变尺度量测获得。本文提出的算法采用逐级匹配的策略，加快了匹配的速度，提高了匹配准确度。实验结果表明这种算法具有很好的适应性。这种方法也可用于其他文字识别中。

This paper proposes an approach to feature points matching for English signature verification. With the comparison of the local characteristics and the relation of the feature points, the similarity between each pair of feature points is calculated and then the matching possibility can be obtained. The feature points set is extracted by the scaled measurement from the stroke skeletons. The approach adopts the hierarchic matching strategy to speed up the matching, and at the same time, the accuration of the matching is improved. The experimental results show that such an approach is robust to the deformation and the noise.It is also able to be used in character recognition.