数字图像的不变特性与特征提取  

Invariant Features of Digital Image and Feature Extraction for Recognition

作　　者：张元[1] 王广宇[2] 吕运朋[3] 

机构地区：[1]郑州大学机电工程系 [2]郑州轻工业学院,郑州450052 [3]郑州大学物理工程学院,郑州450052

出　　处：《郑州大学学报（自然科学版）》2000年第4期57-59,共3页Journal of Zhengzhou University (Natural Science)

基　　金：:河南省杰出青年科学基金资助项目

摘　　要：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.\ \ \}

关 键 词：不变特性 特征提取 模板匹配 字符识别 数字图像 图像变换法 图像识别 最小距离法 方向统计特征量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]