【摘要】 针对目前双弹性网格特征提取方向的重复性和对角弹性网格数量不确定性问题,结合重叠模糊技术提出一种规范化的重叠模糊双弹性网格特征提取方法。对原汉字图像进行外接正方形填补,规范化对角网格,使用弹性技术进行网格划分,在纵横、对角网格中分别构造特征提取方向互不重复的重叠模糊网格,分别提取汉字图像的横竖、撇捺特征。实验结果表明,该方法提取的手写汉字特征类内类间方差比更大,识别率更高。更多还原

【Abstract】 The repeatability of existing feature extraction in elastic dual-mesh and uncertainty of the number of elastic diagonal mesh,an overlapped,fuzzy and standardized elastic dual-mesh was proposed.Diagonal elastic mesh was standardized by filling external square of image.The elastic dual-mesh was obtained through the dynamic meshing.The low relevance’s fuzzy overlapped meshes in the direction of feature extraction were constructed in vertical-horizontal elastic mesh and diagonal mesh,respectively.Chinese characters features were extracted separately based on vertical,horizontal,right-falling,and left-falling stroke.Experimental results in Chinese character recognition show that the feature variance ratio of between classes and within classes is enhanced,and recognition accuracy is also improved.更多还原