【作者】 王虎；

【导师】 吴建国；

【作者基本信息】 安徽大学 ， 计算机应用技术， 2006， 硕士

【摘要】 在我国国民经济信息化迅速发展的今天，传统的计票方式已经远远不能满足社会的需求。以往的民主选举过程中，选票的统计工作采用人工唱票的方式，进行记录和统计，这种方式费时费力，不仅效率低，而且在统计过程中很容易出错，又由于监督不便，很难保证评选工作的公正性，影响了选举工作的进度及质量。“基于图像识别技术的选举计票系统”的设计解决了这一难题，该系统不但适合小型选举，而且适用于具有上千候选人的大型选举，可以对选票进行及时的处理并打印出选举结果，具有高效、公平的特性。整个系统结构简单，成本不高，提高了选票的处理速度和准确性，节省大量的人力和物力，提高了我们的工作效率。 我们设计的“基于图像识别技术的选举计票系统”由图像获取系统、机械传动系统、计算机、网络等构成的光、机、电、信息一体化装置，采用与普通光学标记阅读机完全不同的技术，使用光学字符识别的方法、算法完成选票上标记的识别。即首先通过CCD摄像头把选票作为整幅图像输入，再用算法定位、分割、按照文字的识别方法，识别出“√、×、○、●、＼、／”等各种手写符号，对于包含了另选他人的选票，把包含姓名的图像块分割出来，并显示在计算机屏幕上进行人工校验、录入，大大降低了选票的填写和制作要求，减小了误识率，提高了选票的处理速度。 本论文分为五个部分： 第一部分首先介绍了该课题研究的背景以及国内外选票处理系统的发展状况，接着介绍了整个系统的体系结构。 第二部分介绍了智能票箱的图像获取系统和机械、电子装置的设计与构造，图像数据的采集和传输过程。 第三部分重点讨论了图像灰度化、二值化、倾斜校止、平滑、倾斜纠正、细化以及归一化等选票信息识别中非常重要的图像处理技术。针对选票图像的特性，以提高识别率为目的，提出了恰当的预处理方法。 第四部分介绍了目前主要的字符识别方法，并根据系统的需要，实现了对手写符号的识别。 第五部分详细介绍了计算机网络体系结构和TCP／IP协议，对网络拓扑结构进行了比较，提出一种基于TCP／IP网络编程的方法来解决用一台机器去控制多台智能票箱的问题，实现主控机和票箱之间的数据通信以及功能控制。 最后讨论了设计的不足和改进之处，并对计算机控制的会议选举系统的发展进行了总结和展望。更多还原

【Abstract】 With our country’s economic information developing rapidly, the method of traditional vote counting can’t meet the needs of the society. Difficult problem can be solved by vote processing system based on image recognition, which may be used not only the small-scale vote but also the large-scale vote of possessing thousands of candidates, and the voting result can been printed with high efficiency at the same time. The whole system’s structure is very simple and inexpensive. A great deal of manpower and material sources will been saved by using the system. Then work efficiency will be improved greatly.The system is an integrative device composed of imaging collection, transmission mechanism, computer, networking subsystems. Differs from other vote processing system which is based on the optical mark recognition, the system adopt methods, algorithms of optical character recognition techniques to recognize the handwritten symbols on votes. That is, first, using CCD to input the vote info as a whole image, then using some algorithms to locate, segmentalize characters, then recognize all kinds of handwritten symbols such as "√、╳、○、●、╲、╱". As for the votes which have Other Candidates written in Chinese character, the system segmentalize the characters as small image and display on the computer screen, then the images been verified and input to database. The strategy can greatly reduce the demand of vote production, and improve efficiency of vote processing.This dissertation is divided into five parts altogether:In the first part, the background of the research and the using state of vote processing system through the world are discussed in the dissertation, then, introduced the architecture of the system.In the second part, designing and conformation of image collection system, mechanical and electronic devices been introduced, then, discussed the collection and transmission of image data.In the third part, several important image processing techniques of vote information recognition have been discussed, such as transforming of the color plate to gray plate, image binarization, skew correction, image smoothing, thinning, normalization, et. In更多还原