基于特征行必要-充分性匹配的字符识别方法  

A Character Recognition Approach Based on Feature Line Necessary-Sufficient Condition Detection

作　　者：李佐[1] 王姝华[1] 蔡士杰[1] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,江苏南京210093

出　　处：《软件学报》2002年第1期85-91,共7页Journal of Software

摘　　要：字符识别系统的性能在很大程度上依赖于所选取的字符特征.提出了一种基于特征行必要-充分性匹配的OCR(optical character recognition)方法.该方法使用字符模板的特征行集,通过对待识字符位图进行必要性和充分性双向匹配来识别字符.并采用基线对齐归一化方法在特征匹配时准确定位,使识别率和识别速度都较为理想.同时,对字符位图特征行的交互选择、测试和调整等方法做了详细介绍.另外,还提出了字符骨架与位图重叠显示的方案,有效地提高了对识别结果校对的速度.最后,通过测试和比较,对识别效率进行了分析.The performance of a character recognition system depends heavily on what features are being used. In this paper, an optical character recognition method based on match of feature lines is presented. By extracting the feature lines from bitmap of character and detecting the necessary-sufficient condition with templates by baseline superposed, this method carries out the recognition with a very high efficiency. The process of selecting and adjusting the feature lines from template is also described. Then, a new way for collating is proposed. By the hand of displaying skeletons of recognized characters that overlap the original bitmap, this method makes finding errors easier. Finally, the performance evaluation based on comparison is given.

关 键 词：计算机图形学 字符识别 特征抽取 特征行 必要-充分性匹配 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]