复杂背景下的票据字符类型识别方法

Method of bill character type recognition under complex background

由于银行规定用户需用黑色碳素笔填写票据中信息,但实际操作中存在信息填写不规范现象。针对这种情况,提出一种自动识别用户书写字符类型的算法。该算法首先识别字符颜色,并根据字符在红外图像下是否存在,对字符类型进行粗分类;其次,选择正面红外和正面彩色图像中较清晰的图像作为字符分割的原图像,分别采用阈值法和改进的连通域分析方法对票据正面红外和正面彩色图像中的字符做分割;再次,根据字符笔墨颜色和形体特征,对字符进行颜色判定、基于阈值法的粗细判定、基于位置关系的手写体或打印体判定;最后,设计字符类型识别器,准确识别字符类型。经过实际银行票据图像测试,相比于现有方法,该文采用的方法不损坏票据样本,能够准确地识别出银行票据中的字符类型,具有较好的应用前景和商业价值。

Although the bank stipulates that its users should fill in the bill with a black carbon pen,there are still some non-standard phenomena of information filling in the actual operation. An algorithm for automatic recognition of the character types written by the users is proposed to solve above problems. In the algorithm,the character color is recognized firstly,and then the character type is roughly classified based on whether the character exists in the infrared image. After that,the clearer images are selected from the front infrared and color images as the original images of the character segmentation,and the characters in the front infrared and color images of the bill is segmented by means of the threshold method and the improved connected component analysis method respectively. According to the ink color and physical feature of the characters,the color judgment,the thickness judgment based on the threshold method,and the handwriting or print judgment based on the position relationship are carried out in the character judgment. The character type recognizer is designed to accurately recognize the character type. The testing results of actual bank bill images show that,in comparison with the existing method,the method adopted in this paper does not damage the bill sample,and can accurately identify the character type in the bank bill,which has a good application prospect and commercial value.