Abstract: The characteristics of the automatic recognition of Machine Readable Travel Documents（MRTDs） are analyzed.A detailed design of information recognition method of MRTDs based on the analysis of the characteristic of MRTDs image is given，which consists of four primary procedures：location of machine readable area based on connect component analysis，linear fit based slant correction of machine readable area image，priori knowledge based adaptive segmentation of character image and analysis of structural features based character recognition.Using this method，an automatic information recognition system of visa image is implemented.The test result of the system shows that this recognition method has rather good performance on MRTDs information recognition.

Key words: text location, slant correction, character segmentation, structural feature, pattern recognition

摘要： 分析了旅行证件信息自动识别的特点，提出了一种针对机读旅行证件图像特点的信息自动识别方法，包括基于连通区域分析的机读码区域定位、基于线性拟合的机读码区域图像倾斜校正、基于先验知识约束的自适应字符图像分割以及基于字符结构特征分析的字符识别等四个主要步骤。并在此基础上实现了签证信息自动识别系统。实验表明该方法对旅行证件信息进行自动提取和识别具有较好的效果。

关键词: 文本定位, 倾斜校正, 字符分割, 结构特征, 模式识别