【作者】 杜峰；

【导师】 张国山；

【作者基本信息】 天津大学 ， 控制理论与控制工程， 2007， 硕士

【摘要】 机器视觉技术广泛应用于包装、汽车、机械加工、电子、食品等各个行业。本论文基于机器视觉技术设计了票据喷码质量监控系统的实验装置。整个系统以电子、软件、数字图像处理为基础,能够对喷码票据上的重要标识进行识别。本论文针对票据喷码监控系统设计中所涉及的几个关键技术进行了以下研究工作:1、根据机器视觉技术应用到工业检测上的特点与实际的工业生产现场,设计了一套喷码监控系统的实验装置。该装置的设计内容主要包括三个方面:图像采集设备的选择涉及到工业相机、图像采集卡的选型喷码系统中的光源选择和照明方式设计实验机械装置设计。该装置能较好适应实际生产的需要,可对相机和光源调节进行影响因素分析,元件采用滑动可调的固定方式使调试更容易。2、研究了针对具有复杂底纹背景的票据图像分割方法。利用基于形态学的多尺度边缘检测算法来提取票据图像的边缘,并根据得到的边缘图像梯度直方图自适应的确定图像分割的阈值进行图像分割。3、在本系统中图像识别模块的核心是采用模板匹配识别方法。研究分析了模式识别系统的基本构成,并且比较了模板匹配法和模式识别的几种方法:统计识别方法、句法识别方法、模糊识别方法、人工神经网络法。介绍了基于特征匹配模板匹配方法的优势与特点,本系统采用了以灰度骨架欧拉数特征作为一级分类特征,笔划密度特征、栅格特征作为第二级细化分类的模板匹配方法。4、利用VC++开发环境开发了一套票据喷码监控系统的可视化软件。检测软件的整体规划采用模块化的思想,系统包括图像采集模块、系统调度模块、图像处理模块、通讯功能模块。更多还原

【Abstract】 Machine vision technology is widely used in packaging, automotive, machining, electronics, food and other industries. Note printing quality control system experimental device is designed based on machine vision technology. The whole system is based on electronic technology, software technology, digital image processing technology and the important marks on the note is recognized. In the theses the design of the note printing quality control system involves several key technologies for the following research work:1、According to the characteristics of machine vision technology which is applied to the detection of the industrial production and the actual scene of the industrial production a set of note printing quality control system device is designed. The design of the device mainly includes three aspects: the selection of image acquisition of equipment involved in the selection of industrial camera and image acquisition card; the selection of light source and lighting design ; experimental mechanical design. The device can be suited to the needs of actual production and the impact of the camera and light regulation on system can be analysed. Adjustable sliding components used in fixed mode make debug easier.2、The image segmentation algorithm used in the specific note with complex shading background is proposed. Use the morphology of multiscale edge detection algorithm to extract the image edge and utilize the histogram of the edge of the gradient image to determine the threshold for image segmentation adaptively.3、The template matching recognition method is the core of the image recognition system module. The basic structure of pattern recognition system is analysed and a comparison between the template matching method and several other pattern recognition methods including statistics recognition, syntax recognition, fuzzy recognition, artificial neural network method is drawed. An introduction of the advantages and characteristics of the feature-based template matching method is made. The note printing quality control system uses the Euler feature of the image gray skeleton as the first classification layer and the stroke density feature and grid feature are used as a second-class classification layer.4、Use the VC + + development environment to develop a note printing quality control system visualization software. The design of software is based on the modularization of thinking and the modules include image acquisition module, scheduling module, image processing module, communication module.更多还原