基于KRTS的实时视觉处理系统与模板匹配算法研究

发布时间：2018-08-02 18:56

【摘要】：随着劳动力成本的不断提高以及在以智能制造为核心的工业4.0及中国制造2025战略的不断推进下,国内制造业企业正积极地研制或采购相应的自动化设备对现有生产线进行自动化改造以提高生产效率、降低人工成本、解决制造业劳动力日益短缺等问题。机器视觉是自动化设备的眼睛,通过机器视觉可以进一步提高自动化水平,因此机器视觉近年来被越来越广泛地应用于自动化行业,机器视觉的前景一片光明,但同时也对图像处理的实时性提出越来越高的要求。本文在对机器视觉及其控制系统的国内外研究现状与发展趋势分析的基础上,结合德国Kithara公司开发的Windows操作系统的实时扩展套件KRTS构建了一个实时视觉处理系统并通过实验验证了该视觉处理系统的实时性。通过对串口、USB和以太网等常用通信方式进行优劣势比较之后,本文最终选择综合性能优异的以太网通信,并利用Socket提供的用于实现TCP/IP协议的API函数来实现视觉处理系统与运动控制系统之间的以太网通信。在充分了解Socket实现TCP/IP协议的内部运行机制以及工程中各个变量、数组、结构体等在计算机内存中分布的基础上,本文提出将在视觉处理系统与运动控制系统之间需要进行传输的数据按功能分类并封装成一个个结构体并同其余需要传输的变量一同构成在一个大结构体,Socket再直接利用该结构体对象实现在视觉处理系统与运动控制系统之间的数据传输,省去了数据的序列化与反序列化过程,不仅提高了数据通信效率而且简化了数据传输协议。本文对基于圆投影的传统模板匹配算法进行了改进,通过对工件ROI图像的提取以去除工件外像素对工件识别的干扰;通过对圆形模板制作原则的修改以提高工件信息的利用率;通过粒子群迭代方式,提高了工件旋转定位速度;利用待测工件与模板工件的外部轮廓匹配信息进行重叠工件ROI图像的分离提取;对未完全进入视野的工件提供了 ROI图像补偿机制,以识别该类零件等,使模板匹配结果更加可靠。最后,本文结合OpenCV函数实现并验证了该模板匹配算法,将其移植到本文构建的实时视觉处理系统中。
[Abstract]:With the rising labor cost, and with the advance of industry 4.0 and the strategy of 2025 made in China, with intelligent manufacturing as the core, Domestic manufacturing enterprises are actively developing or purchasing corresponding automation equipment to automate existing production lines in order to improve production efficiency, reduce labor costs and solve the growing shortage of manufacturing labor. Machine vision is the eye of automation equipment, through machine vision can further improve the level of automation, so machine vision has been more and more widely used in the automation industry in recent years, machine vision has a bright future. But at the same time, more and more requirements for real-time image processing are put forward. Based on the analysis of the present situation and development trend of machine vision and its control system at home and abroad, A real time vision processing system is constructed with KRTS, a real time extension suite of Windows operating system developed by Kithara Company of Germany. The real time performance of this vision processing system is verified by experiments. After comparing the advantages and disadvantages of common communication methods such as serial port USB and Ethernet, this paper finally chooses Ethernet communication with excellent comprehensive performance. The Ethernet communication between the visual processing system and the motion control system is realized by using the API function provided by Socket to implement the TCP/IP protocol. On the basis of fully understanding the internal running mechanism of TCP/IP protocol implemented by Socket and the distribution of all variables, arrays, structures and so on in the computer memory in the project, In this paper, it is proposed that the data needed to be transmitted between the visual processing system and the motion control system can be classified according to function and encapsulated into one structure, and together with the other variables that need to be transmitted, it is constructed in a large structure with socket and straight again. Then the structure object is used to realize the data transmission between the visual processing system and the motion control system. The process of data serialization and deserialization is eliminated, which not only improves the efficiency of data communication, but also simplifies the data transfer protocol. In this paper, the traditional template matching algorithm based on circular projection is improved, by extracting the ROI image of the workpiece to remove the interference of the outside pixels of the workpiece to the identification of the workpiece, and by modifying the principle of making the circular template to improve the utilization rate of the workpiece information. Particle swarm optimization (PSO) is used to improve the rotation speed of the workpiece, and the ROI image of the overlapped workpiece is extracted by using the matching information of the external contour of the workpiece to be tested and the template workpiece. The compensation mechanism of ROI image is provided for the workpiece which is not fully in view, so as to identify the parts, and make the matching result of template more reliable. Finally, this paper implements and verifies the template matching algorithm with OpenCV function, and transplants it to the real-time visual processing system constructed in this paper.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】