针对大形变增值税发票汉信码的识别算法

发布时间：2018-01-01 08:02

本文关键词：针对大形变增值税发票汉信码的识别算法　出处：《吉林大学》2016年硕士论文　论文类型：学位论文

【摘要】：进入二十一世纪以来,二维条码已然成为信息化时代的新宠,遍布于日常生活的每个角落,是人们衣食起居不可或缺的重要角色。汉信码作为二维条码的代表码制更广泛地应用于增值税发票信息的存储、传递和识别。糟糕的是,受限于打印机和扫描仪等硬件设备的磨损老化问题,以及图像采集过程中的诸多不可控因素,发票上汉信码的图像质量参差不齐,严重影响信息的正确识别与传递。诚然传统的发票汉信码的识别算法思路清晰、易于实现,且对于图像质量较高的发票码图已经达到令人满意的识别率,但它对于发生较大不规则形变的发票码图却束手无策。而近年来恰恰是此等图像质量较低的发票于市场中大批量地涌现,使得人们逐渐认识到解决大形变的发票汉信码的识别问题已刻不容缓。基于此现状,本文针对大形变的增值税发票汉信码,提出了一种新的识别算法。首先,将传统的识别流程模块化,分析影响识别率的问题所在,并保留其合理模块;其次,将码图大形变分为纵向整体形变与横向不规则形变,二者区别对待,依次解决。利用灰度图像进行固定信息的跟踪,利用改进的二值图像进行黑白模块交界的确定。最后,按照抽屉原理的思想制定合理的“生成准则”,依据得到的上述图像数据,生成对应于原大形变码图的标准码图,而后直接译码即可获得发票的真实信息。为证实本文算法的优势,特将实验结果一并附上。对于传统算法只能达到20%识别率的大形变发票,本文算法的识别率在80%以上。
[Abstract]:Since 21th century, two-dimensional bar code has become the new favorite of the information age, all over the daily life in every corner. Han letter code, as the representative of two-dimensional bar code, is more widely used in the storage, transmission and recognition of VAT invoice information. Limited by the problem of wear and aging of hardware equipment such as printer and scanner, and many uncontrollable factors in the process of image acquisition, the image quality of Han Xin code on invoice is not uniform. It seriously affects the correct identification and transmission of information. It is true that the traditional invoice code recognition algorithm is clear, easy to implement, and the image quality of the invoice code map has reached a satisfactory recognition rate. However, it has no way to deal with the large irregular deformation of invoice codes, and in recent years, the invoices with low image quality are emerging in large quantities in the market. It makes people realize that it is urgent to solve the problem of large deformation invoice code recognition. Based on this situation, this paper proposes a new recognition algorithm for large deformation value added tax invoice and Han letter code. The traditional identification process is modularized to analyze the problems affecting the recognition rate and to retain its reasonable module. Secondly, the large deformation of code map is divided into longitudinal integral deformation and transverse irregular deformation, which are treated differently and solved sequentially. The fixed information is tracked by gray image. Using the improved binary image to determine the boundary of the black and white module. Finally, according to the principle of drawer to formulate a reasonable "generation criteria", according to the above image data. The standard code graph corresponding to the original large deformation code graph is generated, and the true information of the invoice can be obtained by direct decoding. The experimental results are attached. For the large deformation invoice which can only reach the recognition rate of 20%, the recognition rate of this algorithm is more than 80%.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】