基于机器视觉的机器人二维曲线静态跟踪

【作者】 王伟；

【导师】 林家恒；

【作者基本信息】 山东大学 ， 模式识别与智能系统， 2005， 硕士

【摘要】 对金属材料的切割和焊接是工业生产中的一道重要工序,广泛应用于机械制造、建筑、电力、水利、造船等领域。目前在国内,主要还是由人或传统生产设备来完成,生产效率较低。如果能采用具有机器视觉的机器人自动完成,那么不仅可以提高产品的质量和产量,还可以增强我国企业的国际竞争力,机器人对金属材料的切割和焊接抽象出来可以归结为机器人对加工曲线的静态跟踪。因此针对机器人加工曲线视觉跟踪问题进行深入的研究具有较大的工程实用价值。 论文综合应用了机器视觉,CMAC神经网络和机器人控制技术,设计并实现了基于机器视觉的机器人二维曲线静态跟踪系统。论文的研究内容主要包括跟踪曲线的图像处理、CMAC神经网络坐标变换和机器人跟踪控制。 1.跟踪曲线的图像处理:分为预处理、二值化、二次处理、细化和曲线提取这几个过程。应用了图像灰度化,中值滤波,线性变换,阈值二值化,串行迭代细化等算法,用Visual C++作为开发工具,实现了对跟踪曲线的提取。 2.CMAC神经网络坐标变换:设计了CMAC神经网络,实现了从跟踪曲线的图像二维坐标到机器人内部六维坐标的变换,设计了训练样本的采集方案,采集了足够的样本。采用了改进的CMAC网络,明显提高了网络的映射精度。进行了大量的仿真和现场坐标变换实验。该变换是用MATLAB工具编制的。 3.机器人跟踪控制:对SK—6型机器人控制系统进行了消化、试验和研究,通过VB编写了上下位机的通讯程序。综合使用了机器人的三种控制方式:示教编程、远程指令、机器人语言编程,实现了机器人沿跟踪曲线等时间间隔点点折线移动,完成了二维曲线的静态跟踪。 总之,本文初步解决了机器人对平面曲线的静态跟踪问题,展示了机器人在工业生产中的良好应用前景。更多还原

【Abstract】 The cutting and welding of metal material is an important process in industry and it is widely used in fields such as machinery manufacture,architecture,electricity, water conservancy,shipbuilding,etc.Now in our country it is mainly finished by people or traditional production equipments and the production efficiency is relatively low .If the job can be automatically finished by robots based on machine vision,not only the production’s quality and outputs can be improved rapidly ,but also our country’s enteprises’ international competitiveness can be strengthened greatly. The job that robots cut and weld the metal material can be abstracted into robots’ static tracing for the object curve. So deep research about robot’s curve tracing has greater practical value.The thesis comprehensively applies the machine vision technology, CMAC neural network technology and robots’ control technology; designs and realizes the static tracing system of robots based on machine vision for 2-D(dimensional) curve. It mainly includes object curve’s image process, CMAC ’s coordinate transformation and robots tracing control1.Object curve’s image process: This part is divided into steps like preprocess,binarization,second-process, thinning and curve extraction. Medium filtering, linear transformation, iterative and other algorithms are used in this part .The processing program are compiled in Visual C++ .2.Transformation of coordinates through the CMAC neural network: This part designed the CMAC neural network, realized the transformation from the object curve’s 2-D coordinate to robot’s 6-D coordinate, collected enough training samples, used a new kind of CMAC ,improved the network’s precision obviously and carried out a lot of experiments.MATLAB tools is used in this part.3.Robot tracing control:In this part we digested ,tested and studied the SK-6 robot.,compiled the communiction program between PC and the robot,used therobot’s three control mode and realized robot’s 2-D curve static tracing .In conclusion the thesis primarily resolved the robot’s tracing for two-dimentional curve and the bright future of robots’application in industry is foreseeable.更多还原