【作者】 庞然；

【导师】 汪仁煌；

【作者基本信息】 广东工业大学 ， 检测技术与自动化装置， 2008， 硕士

【摘要】 视觉检测与识别技术已经逐渐深入到工业的各个领域中,特别是在工业流水线上的视觉检测技术,极大地提高了制造业的自动化水平。液晶数字表以其特有的优势逐渐应用到如汽车行驶参数显示、带有液晶屏的嵌入式电子产品、消费类电子、测量仪器等领域中。液晶显示的出现,给信息的获取提供便利,如能对其显示的信息自动提取,以便反馈给执行系统,则会在更大程度上提高信息获取的效率。在该背景下本文探讨研制了一台能自动识别液晶数字表的机器视觉系统,主要创新点和完成的工作有:1、对摄像机、光源的选择进行了有益的探讨在图像获取方面,合理选择摄像机和光源,是视觉图像检测识别的前提。本文在分析各种类型工业用摄像机的基础上,得出有关摄像机参数的选择,如:分辨率、运动状态下的扫描速率等。讨论了同轴光源在消除液晶显示时背景反光的干扰作用,为获取清晰的图像奠定了基础。2、去除液晶数字表图像背景大面积噪声干扰、二值化过程中的孤立噪声点等关键问题。本文在国内外液晶数字表检测识别已有研究成果的基础上,重点讨论以下关键问题:解决图像大面积背景噪声干扰和二值化过程中的孤立噪声干扰等问题;在一般意义的同态滤波基础上改进同态滤波函数的参数,设计新的滤波函数,能消除液晶背景特有的阴影暗区干扰。此外,用最佳二值化阈值处理图像后出现的孤立噪声点,会混入到二值化后的图像中,干扰有效的数字信息识别,本文设计了去除孤立噪声和数字粘连噪声的算法,并列出效果图,有效地去除了噪声的干扰。3、设计了一套液晶数字表识别模拟系统,丰富了基于Matlab图像处理平台的识别对象和函数,拓展了其应用范围。扩展基于Matlab图像处理软件平台的功能,在最初的框架模式下,本文的一些研究成果成功地嵌入到了该平台下。主要算法,如同态滤波、去除孤立点噪声和粘连噪声、数字孔洞的填充等都能针对实时显示的图像进行处理,并得到清晰、易于识别的二值图像,解决了数字识别前图像预处理中的关键问题。同时,为了验证算法和系统的有效性,设计了模拟识别系统,采用YM12864R液晶模块显示数字,并用51MCU和PC软件平台进行通信,在这一平台下仿真,取得的效果具有一定的实用价值。更多还原

【Abstract】 Visual Inspection Technology has been gradually applied to industrial fields, especially the technology used in the industrial pipeline, which have greatly improved the level of automation in the manufacturing field. Numeral LCD instrument have been gradually applied to many areas, such as the traffic parameters display of motor vehicles, embedded electronic products with LCD screen, consumer electronics, measuring instruments by its unique advantages. The emergence of liquid crystal display facilitates acquirement of information. Equipment that can automatically extract these information and bring feedback to the implementation of systems, would greatly improve efficiency of the information attainment. For these reasons, the paper tries to research and develop a machine vision system that can automatically recognize the number in LCD. The main achievement can summarized as follows:1、Develop a instructive discuss to the camera and lightingIn the part of image acquisition, choose a suitable camera and lighting is the key to the image recognition. This paper focus on how to calculate those parameters such as resolution and scan rate in the state of movement, also it discus how to eliminate noise brought by glisten after using the coaxial lighting. All these discussions lay the foundation for attaining a clear image.2、Remove large background noise of the LCD image, isolated noise existed in the process of binary image.Based on the research of LCD image recognition at home and abroad, this thesis study on main question is as followed. Removing of Large background noise of the LCD image, isolated noise existed in the process of binary image. Through modifying some parameters, improve filtering function base on homomorphism filter so that eliminating large background noise. As much isolated noise will exist in the binary image after image processing, and interfere the recognition. This thesis develops algorithm to eliminate these noise, and give the result after processing.3、Design a simulation system of LCD recognition, develop new functions to expand the scope of the appiication of the platform based on Matlab.Base on the original software structure, achievement of this thesis can embed to the image processing platform. These algorithms includes homomorphism filter, removing of isolated noise and conglutination noise, holes filled and so on. All these method can process the real image and acquire a clear binary image which is the key point in the preprocessing. To verify the effectiveness of algorithms and system, designing a simulation system base on single chip is needed. This system is made up of LCD module YM12864R and C51 microchip processor, and attaining practical value.更多还原