基于印刷体汉字识别的快递邮包分拣系统

发布时间：2018-06-15 18:28

本文选题：汉字识别 + 字符识别　；参考：《合肥工业大学》2017年硕士论文

【摘要】：近年来,我国快递服务业高速发展,信息化水平不断提高,大量的基于条形码和二维码的包裹自动分拣系统开始应用到分拣过程中。然而,快递分拣现场环境复杂,条码污损的情况时有发生。针对条码无法读取的包裹,快递公司通常采用人工识别收发地址字符的方式实现分拣,成本较高且效率低下。基于图像处理的汉字识别技术具有速度快,成本低,自动化程度高等优势,在快递邮包分拣系统中拥有广阔的应用前景。论文首先介绍了分拣系统的整体架构和工作原理,硬件平台的设计和选型,并简要叙述了软件平台的设计。接着根据快递单图像的高亮度特征,使用多阈值大津法、形态学操作和连通域筛选得到快递单位置,并利用条码区域的高对比度特征设计了补充定位方案。然后通过霍夫变换对图像角度进行了修正,在多次分析投影波形后找到字符位置,并对字符图像进行了拆分和归一化。在字符识别阶段,首先叙述了样本字符数据库的制作过程,然后利用像素网格特征和梯度方向网格特征对字符图像进行特征提取,最后通过标准欧式距离分类器实现了字符的识别。论文最后展示了邮包分拣系统硬件环境搭建结果和软件平台开发结果,并用大量的包裹进行了验证。实验结果表明本系统字符识别准确率达到99.76%,包裹分拣准确率达到98.35%,图像处理耗时约为0.5秒,满足快递包裹分拣环节的要求。使用本套系统,可以降低人力成本,提高分拣效率,或和条码识别技术结合,提高分拣系统的冗余性。
[Abstract]:In recent years, with the rapid development of express service industry in China, the level of information has been improved, a large number of parcels automatic sorting system based on bar code and two-dimensional code began to be applied to the sorting process. However, the scene of express sorting complex environment, bar code fouling occurred from time to time. For parcels that can not be read by barcode, express delivery companies usually use manual identification to send and receive address characters to achieve sorting, which is costly and inefficient. The Chinese character recognition technology based on image processing has the advantages of high speed, low cost and high degree of automation, so it has a broad application prospect in the express mail packet sorting system. This paper first introduces the whole structure and working principle of sorting system, the design and selection of hardware platform, and briefly describes the design of software platform. Then according to the high luminance feature of express single image, we use the method of multi-threshold, morphological operation and connected domain screening to get the single location of express delivery, and use the high contrast feature of bar code region to design a supplementary location scheme. Then the angle of the image is modified by Hough transform, and the character position is found after analyzing the projection waveform many times, and the character image is split and normalized. At the stage of character recognition, the process of making sample character database is described, and then the feature extraction of character image is carried out by using pixel grid feature and gradient direction grid feature. Finally, the character recognition is realized by the standard Euclidean distance classifier. At the end of the paper, the results of hardware environment and software platform development of packet sorting system are presented and verified by a large number of parcels. The experimental results show that the accuracy of character recognition is 99.76 and the accuracy of package sorting is 98.355.The processing time of image is about 0.5 seconds, which meets the requirements of the sorting link of express package. The system can reduce the labor cost, improve the sorting efficiency, or combine with the bar code recognition technology to improve the redundancy of the sorting system.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F259.2;TP391.4

【相似文献】