数字图像的不变特性与特征提取

Invariant Features of Digital Image and Feature Extraction for Recognition

讨论了两种字符特征向量的提取方法 .一种是基于平移、旋转和尺度不变性的图像变换法 ,另一种是强调字符形状和结构的方向特征量提取法 .比较试验的结果表明 ,通过计算字符图像中心矩的图像变换法具有比较稳定的特征值 ,有利于机器识别 .实验中通过最小距离法进行字符的分类识别 ,识别率达到 80 %以上 。

Introducing two methods of feature extraction, one is based on image transformation \{with the invariant\} features which keep stable while the picture move, rotate and zoom in or out, \{and the other\} is focused on the shape of character and features of directin. Experiment results indicate \{that stable features\} of character can be obtained, and it is useful for recognition. Moreover, the \{minimum distance matching\} algorithm used in the recognition system of character is reliable and \{efficient. The\} rate of recognition is about 80%. It can be used in vehicle license plate recognition \{system.\ \ \}