【作者】 张巍；

【导师】 陈海鹏；

【作者基本信息】 吉林大学 ， 软件工程（专业学位）， 2015， 硕士

【摘要】 随着现代科学技术的发展，移动智能设备已经越来越深入人们的生活。从最初人们使用手机进行简单的通话，发信息，到如今人们开始使用手机听音乐看电影，甚至进行文本编辑，图像处理等等各种各样的以前难以想象的操作。在现代的商务以及社会活动中，名片作为一个重要的代表身份信息的载体，应用愈加广泛。随着社会信息化程度不断进步，大量身份信息的识别与管理称为迫切需求。按照传统的方式，人们需要名片夹保存生活中收到的诸多名片，在查找时也需要一张一张的查看，非常不方便。结合应用日益广泛的移动智能设备，如果能利用这些智能设备自动完成名片的整理收集，一定会在很大程度上给人们的商务生活带来便利。因此对基于移动终端的名片识别管理系统开发与研究是具有实际意义的。本文就是在此基础上设计开发一款基于android平台的名片扫描识别系统。而在日常生活中根据语言类型分类不同，大致有单语名片和混合多语名片，在本文中以英文名片的图像分割和识别为主要研究课题，说明了其相关算法和现实意义。基于当下最为普及与流行的移动端操作系统Android平台开发系统，使工作更具现实意义。如何对名片中的文字信息进行良好的提取和识别，是本文首先需要考虑的问题。而OCR技术作为一种相对成熟、准确的文本识别技术，非常值得深入研究。本文在信息识别方面主要采用OCR文字识别技术。在决定一个OCR库之前，开发者首先需要决定OCR过程应该发生在智能手机上，或者是在云端。每种方法因其适用于各自环境而具有相对的优越性。本程序主要实现的是TESSERACT OCR引擎在Android平台上的移植。它由HP于1985年到1995年间开发，后来由Google直接负责，在进一步的开发之后，有了显著的改进。TESSERACT OCR和Leptonica图像库协同工作，兼容多种图像格式，可将图像转换成超过60种语言的文本。除了可以在Android和Iphone平台上编译外，还能工作在Linux，Windows，Mac OSX等系统上。本文首先就Android系统的普遍行以及当前在文字信息识别方面的主要发展趋势与阶段性成果做以介绍。之后主要介绍了OCR图像识别，以及Android系统的历史及现状，并且简要介绍了开发环境及开发工具JDK，Android SDK及Eclipse，对本程序实现的编程语言——Java语言进行了对其相关编程技术的叙述，同时也从Android架构方面入手，简要说明了Android的相关技术及结构。在算法部分，本文介绍了图像的预处理、图像的页面设计、图像分割、图像的字符识别以及后处理，对相关算法进行了说明和比较，着重介绍了图像的字符分割和字符识别。对字符进行处理是系统的重要环节，好的字符处理过程是实现应用功能和提高应用的用户体验的关键所在。在我们的系统中，我们使用两次通过的方法进行字符识别，第一次识别产生基本模式，第二次通过方式被称为“自适应识别”，可更好地认识第一次剩余的信息，输出高可信度的认可字母形状。这对其中的字体如扭曲的（模糊或褪色）不寻常的字体或低质量的扫描是有利的。在android系统下如何进行应用软件的开发也是现在非常热点的课题，开发模式、系统架构都在很大程度上决定了应用系统最终成果的优劣。文中同时还对整个系统的架构与各个模块进行了详细的介绍，并依次给出各自函数。对于用户交互的界面我们着重考虑到用户使用的便捷性，对信息处理与识别的部分我们做了详细的介绍其中对图像捕获，预处理，分割，识别等步骤的程序也给出了解释。一款好的应用程序不仅在开发过程中需要考量各种各样的问题，并且给出好的解决方法。在程序完成之后，也必须有好的程序测试过程来保证应用可以满足设计之初的各种需求，找出程序开发过程中的漏洞和不足，并根据测试结果对程序进行完善，最终得到一款尽可能完美的应用程序。本文也从环境测试和功能测试两个方面对应用程序进行系统测试，保证了开发出的应用系统在各方面符合设计之初的需求并且对用户友好。在经过理论基础研究，程序架构设计，模块划分，具体实现以及系统测试等各个步骤的工作之后，本文最终实现了最初的构想，完成了一款基于android系统的名片扫描识别系统。该系统能够成功识别文本图像中的字母和数字。系统在设计中主要考虑到用户使用体验，针对每个功能设计了相应的按键响应，从拍照捕获图片，到图片的存储，再到对本地图片的搜索比较各个环节都设置了独立的响应，方便用户按照需求选择。鉴于个人时间精力以及个人能力的限制，本文设计实现的基于android系统的名片扫描识别系统很难在各方面做到完美，一定存在其不足或者需要改进的地方。因此在最后本文根据实验结果对研究做了相应评估，对本文开发的应用系统进行总结。同时结合研究中观察到的其他应用程序的特点以及实现方式提出了在研究中存在的问题以及可能的解决方法。于此同时，也提出了符合实际的对应用系统进行进一步优化和方向，并对改进可能做了合理设想。在所有工作基础上就该方向的发展前景给出了合理化意见。更多还原

【Abstract】 With the development of modern science and technology, mobile smart devices hasbeen increasingly deepened in people’s lives. From the original people use mobile phonesfor simple calls, send messages, but now people are beginning to use the phone to listen tomusic, watching movies, or even text editing, image processing, and so a variety ofpreviously unimaginable operation. In the nowadays business and social activities, as animportant representative of the carrier of information, business cards is widely used. Alongwith the social degree of information progressing, a large number of identity recognitionand management is called a urgent demand. According to the traditional way, people needto use the contacts save many received cards, when looking for a view is also veryinconvenient. Combined with the increasingly widespread use of mobile smart devices, ifwe can use these smart devices automatically complete finishing business card collectionwill be in large part to people’s business life a lot easier. So for business card recognitionbased on mobile terminal management is meaningful to research and development. In dailylife, based on the classification of different types of languages， there are basicallybilingual business cards and one language cards. In this paper，business cards which are inEnglish image segmentation and recognition as the main research topics，illustrate thecorrelation algorithm and practical significance.Extraction and recognition of the text of the card information is the problem of thispaper to consider. The OCR technology as a relatively mature and accurate text recognitiontechnology is very worthy of further study. This article mainly uses the OCR characterrecognition technology in information recognition. Before deciding to an OCR library，developers first need to decide OCR process should take place on smart phones, or in thecloud. Each method because of its applicable to their environment and has the relativesuperiority. The main procedures are TESSERACT OCR engine on the Android platformmigration. It is developed by HP from1985to1985, then by Google was directlyresponsible for, after further development, there has been a significantimprovement.TESSERACT OCR and Leptonica image library work, is compatible with avariety of image formats, can transform the image into the text of the more than60languages.In addition to outside the Android and Iphone compilation, can still work onLinux, Windows, Mac OSX systems, such as.First of all, this paper introduce the general line of Android system and the main trends of development and achievements of the current in the aspect of text informationidentification. This paper describes OCR image recognition，and the history and status ofthe Android system, and development tools JDK, Android SDK, Eclipse brief introductionto the programming language of the program implementation-Java language programminghas been described in its associated technologies, At the same time there is a briefdescription of the relevant technical and structure，based on the Android architecture.In the algorithm parts, the paper describes the character segmentation imagepreprocessing, image layout，image character recognition and post-processing, algorithmsare described and compared. Image segmentation and character recognition are highlighted.To deal with character is an important part of the system, good character applicationprocess is to realize the function and the key to improve the application of the userexperience. In our system, we use two times through the method of character recognition,the recognition for the first time to produce the basic pattern, the second time through theway called "adaptive identification", to better understand the rest of the information for thefirst time, the output high credibility approval letter form. The font of them such asdistortion (fuzzy or fade) unusual font is good or low quality of the scan.How to application software development in android system is now very hot topic,development mode, the system architecture are largely determine the final outcome of themerits of the application system. In this paper，we also introduce the whole systemarchitecture and modules in detail, at the same time show their respective function in turn.We mainly consider the user interface for user interaction using convenience, forinformation processing and recognition part we made a detailed introduction of imagecapture, preprocessing, segmentation, identification steps such as the explanation is alsogiven. Meanwhile it is cited in the text important part of the function (keys function，etc.).A good application not only in the development process need to consider a variety ofissues, and gives a good solution. After the process is complete, you must also have a goodtesting process to ensure that applications meet the needs of design at the beginning, theprogram development process to identify vulnerabilities and weaknesses, and based on thetest results of the program to improve and eventually get a as the perfect application. Thisarticle from environmental testing and functional testing two aspects of the applicationsystem test to ensure that the developed application systems designed to meet the needs inall aspects of the beginning and user-friendly.After the theory of basic research, application architecture design, module division, implementation, and operation of the system specific testing each step, the paper finallyrealized the original idea, completed the business card scanning identification system basedon android system. This system can successfully identify the letters and numbers in the textimage. System mainly considering the user experience in the design, the correspondingbutton for each function design response, from the photo captured images, the images ofstorage, and then to search to compare each link of local image set up independent response,convenient user select as needed.Given the limitations of personal time and energy as well as individual ability, thispaper designed and implemented based on the android system identification card scanningsystem is difficult to be perfect in every respect, there must deficiencies or areas forimprovement. At the end of this paper makes the corresponding evaluation according to theresults of research, and put forward the problems in the study, at the same time do thereasonable assumption to possibly improve. On the basis of all the work on the direction ofthe development prospects of rationalization suggestions are given.更多还原