【摘要】 手写体汉字识别一直是模式识别研究领域中的难点。本文将矩变换与弹性网格方法结合起来,用正交Zernike矩提取全局特征,用弹性网格提取局部特征,然后用K-L变换对局部特征向量正交化。这样得到的特征向量不仅具有汉字的全局和局部特征,而且是最有分辨能力的特征向量。实验证明该方法是行之有效的。更多还原

【Abstract】 In this paper,a feature extraction method based on orthogonal Zernike moments and elastic mesh tech-nique is proposed to extract global and local features. Then,K--L transformation is used to compress the local fea-ture. The fusion feature not only presents character’s global and local features,but also is the most discriminativefeature. This method has been proved to be effective.更多还原