基于机器视觉的日化用品泵头缺陷高速在线检测系统研究

发布时间：2018-07-22 16:19

【摘要】：机器视觉技术从发展初期开始便被广泛应用于各个领域,关键在于其速度快、信息量大、功能多等特点。其中,应用最多的是工业生产线上的产品自动化检测领域。采用机器视觉检测系统可以提高检测速度、检测效率,降低工业生产成本,因此近年来对机器视觉技术的研究很多都集中在工业检测领域上。本文根据项目要求,研究了基于机器视觉技术的日化用品泵头缺陷检测系统以及一种新型的ROI区域仿射变换参量的快速获取算法。本文主要研究内容如下:在本文中,第一章综述了本课题的背景、机器视觉技术的发展状况、国内外研究现状、以及研究内容和研究意义。第二章研究了系统需求、总体设计方案以及各功能模块。本章中详细研究机器视觉硬件系统部分内容,机器视觉技术研发离不开相机光源等硬件,并且硬件的选型和光路的设计直接决定着系统的性能以及可靠性,所以还对相机、镜头、光源等各型号以及其适用场合作了详细研究。特别是光源部分,作为检测系统中最为重要的部分,合适的光源以及光路系统直接影响拍摄图像的质量和后期图像处理的难度。所以选择合适的光源跟光路系统可以降低后期图像处理的难度从而增加系统的性能和稳定性。本章还简要研究了眩光如何消除以及眩光的消除原理。第三章主要研究角点检测算法中三种常用的角点检测算子,并深入研究了它们的原理。我们通过对比分析Moravec算子、Harris算子以及SUSAN算子,发现Harris算子对本项目中的待测物进行角点检测能达到最优的效果,所以最后确定使用Harris算子作为本项目图像处理算法的角点检测算法。第四章主要研究角点匹配算法,简要研究角点匹配算法的四要素,并详细研究本人自己推导的一组最优仿射变换参量计算公式(一种新型的ROI区域仿射变换参量的快速获取算法)、关键推导过程及实验验证。通过使用该组公式实现本项目图像处理算法中的角点匹配算法部分,直接快速求出角点匹配所需的平移旋转参量。经实验验证该公式能够非常完美配合系统中的图像处理算法。针对目前研发机器视觉检测系统需要对特定需求开发特定系统,本文第五章研究了基于VC6.0平台的MFC+halcon软件开发的机器视觉软件系统,具体为软件系统的整体结构设计以及功能模块设计。整体结构设计中对相机的初始化、调用回调函数、设置启动定时器、开辟多线程、控制电磁阀等进行了研究。各个功能模块部分主要包括相机控制模块、图像采集模块、图像处理模块、串口通讯、光源控制模块等,在这部分还研究了本系统所用的一部分图像处理算法,图像处理算法在机器视觉检测系统中是最为重要的,也是本项目最为核心的一部分。本研究设计的日化用品泵头缺陷检测系统经过大量的实验测试表明,该系统能对日化用品泵头表面缺陷能准确检测出来,有效实现了日化用品泵头表面缺陷的实时非接触检测。
[Abstract]:The machine vision technology has been widely used in various fields since its early development. The key lies in its fast speed, large amount of information and many functions. Among them, the application of the machine vision is the most widely used in the field of automatic inspection of the products on the industrial production line. The use of machine vision detection system can improve the speed of detection, detection efficiency and reduce the cost of industrial production. So many research on machine vision technology in recent years are concentrated in the field of industrial testing. In this paper, based on the requirements of the project, this paper studies the pump head defect detection system based on machine vision technology and a new fast acquisition algorithm for the ROI regional affine transformation parameters. The main contents of this paper are as follows: in this paper, The chapter summarizes the background of this topic, the development of machine vision technology, the current research status at home and abroad, and the research content and research significance. The second chapter studies the system requirements, the overall design scheme and the functional modules. In this chapter, the part of the machine vision hardware system is studied in detail, and the research of machine vision technology can not be separated from the camera light source. Such as hardware, and the selection of hardware and the design of optical path directly determines the performance and reliability of the system, so the cooperation of the camera, lens, light source and other models, as well as its applicable field, is also studied in detail. Especially the light source part, as the most important part of the detection system, the appropriate light source and optical path system directly affect the picture. The selection of the appropriate light source and optical path system can reduce the difficulty of the later image processing and increase the performance and stability of the system. In this chapter, the principle of how to eliminate glare and the elimination of glare is briefly studied. The third chapter mainly studies the three common corner points in the corner detection algorithm. By comparing and analyzing the Moravec operator, the Harris operator and the SUSAN operator, we find that the Harris operator can detect the corner point of the item in this project to achieve the best effect. Finally, the Harris operator is used as the corner detection algorithm for the image processing algorithm of this project. The four chapter mainly studies the corner matching algorithm, briefly studies the four elements of the corner matching algorithm, and studies in detail a set of optimal affine transformation parameters calculation formula derived by myself (a new fast acquisition algorithm for the ROI area affine transformation parameter), the key derivation process and experimental verification. By using this set of formulas to realize the project The corner matching algorithm part of the image processing algorithm is used to quickly find the translational rotation parameters needed for the corner matching. It is proved that the formula can perfectly match the image processing algorithms in the system. The fifth chapter of this paper is based on the study of V based on the specific requirements of the machine vision detection system. The machine vision software system developed by MFC+halcon software of C6.0 platform is specifically designed for the overall structure design and function module of the software system. The initialization of the camera, the call back function, the setting of the start timer, the opening of the multi thread and the control of the solenoid valve are studied in the whole structure design. It includes the camera control module, the image acquisition module, the image processing module, the serial port communication and the light source control module. In this part, a part of the image processing algorithm used in this system is also studied. The image processing algorithm is the most important part of the machine vision detection system, and it is also the most important part of this project. The test system of the product pump head defect test shows that the system can accurately detect the surface defects of the pump head of the daily products and effectively realize the real-time non-contact detection of the surface defects of the daily product pump head.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】