基于特征融合的维吾尔文笔迹鉴别

针对采用纹理方法鉴别维吾尔文不稳定的问题,提出一种与文本无关、特征融合的笔迹鉴别方法，融合的特征包括网格窗口微结构特征和笔迹曲向特征。所提方法从笔迹原始图像提取笔画边缘，对笔迹的边缘图像建立大量局部窗口模型，通过扫描边缘图像获取融合特征结构的概率密度分布，使用多种距离公式计算概率密度向量间的距离。在实验笔迹容量大小为80的笔迹库上进行实验得到的鉴别率为89.2%。所提方法能很好地刻画笔迹的局部书写变化趋势和笔画的曲向，采用概率密度分布来统计笔迹的网格窗口微结构特征和曲向特征，鉴别效果达到了预期值。

Uighur handwriting identification based on feature fusion

Concerning the instability of Uighur handwriting identification by texture, the authors proposed a text-independent method of handwriting identification based on feature fusion, and feature fusion involved mesh-window microstructure feature and curvature-direction feature. On the basis of extracting edge strokes from original image, a large number of local window models were created. By scanning the edge image, the probability density distribution of the feature fusion structure was obtained. And a variety of distance formulas were used to calculate the distance between the probability density feature vectors. The experimental identification rate is 89.2% in the database involving 80 handwritings. This method can portray the local writing trends of the handwritings and the curvature-direction of the strokes, the proposed method adopts probability density distribution to statistically record the mesh-window microstructure features and the curvature-direction features, and the identification effect is satisfactory.