【作者】 王凯；

【导师】 万小金；

【作者基本信息】 武汉理工大学 ， 动力机械及工程， 2019， 硕士

【摘要】 随着工业制造智能化的概念被提出,以此为目标的工业革命开始普及全球。在传统的自动化生产系统中,将机器视觉系统同工业机器人相结合,使工业生产系统更加智能化,如今已经是一个研究热点。文章中对生产制造产业中使用的工业机器人智能抓取系统进行研究,了解当前这一系统的发展以及关注焦点,分析研究相机在工业机器人上的应用,之后学习了图像的预处理、识别与定位以及机械臂运动学等关键技术,最后以ABB工业机器人作为研究对象,搭建了一个以机器视觉为基础的抓取机器人系统。课题研究中所包含的主要内容如下:首先,对传统Canny算法进行改进。在研究机器视觉的过程中,较为重要的部分就是图像处理这一过程,本文学习了图像预处理技术,分析比较了四种滤波器的优点和缺点;在研究如何提取图像边缘信息时,针对传统Canny算子存在的不足之处,本文提出了改进方案并利用实验进行验证。然后,针对目标识别、定位问题,采用SURF算法完成目标识别工作,并对单目视觉的定位原理以及相机的标定方法进行了描述。采用张正友标定方法对实验所用的相机完成标定,并获取它的内部和外部参数。单目定位采用了特征匹配的方法,通过分解本质矩阵计算得到目标的深度信息,并提取出目标的质心坐标。之后,根据工业机器人机械臂运动学的相关原理,使用D-H法完成机器人的建模工作,并使用ADAMS软件进行机械臂的运动学仿真实验,用来验证所建立的运动学正解模型的准确性。采用了一种软竞争ART-RBF学习算法来对机械臂运动学逆解进行计算。为清晰地了解实验所用的机器的作业范围,使用MATLAB求得机械臂的三维可达空间。最后,根据现有设备资源,建立了视觉引导的工业机器人系统,完成了所选模型的抓取测试。本文通过搭建一个基于视觉引导的工业机器人完成抓取以及放置等动作,学习、研究了机器视觉系统以及工业机器人系统相关内容。通过实验证明了所搭建系统的可行性,为工业生产中使用视觉引导的工业机器人系统提供了一定的参考。更多还原

【Abstract】 With the concept of industrial manufacturing intellectualization being put forward,the industrial revolution aiming at this goal began to spread all over the world.In the traditional automatic production system,the introduction of machine vision to make the production system more intelligent has become a research center point.This paper studies the intelligent grasping system of industrial robots in industrial production,understands the development and research focus of the current intelligent grasping system of industrial robots,mainly studies the application of machine vision in industrial robots,and studies some key technologies such as image preprocessing,recognition and positioning,and kinematics of robotic arms.Finally,ABB industrial machine is used as the research object.Based on this,an industrial robot grasping system based on machine vision has been built.In the course of the research,it mainly includes the following contents:Firstly,the traditional Canny algorithm is improved.Image processing is an important part of machine vision.This paper studies image preprocessing technology.The advantages and disadvantages of Gauss filter,mean filter,median filter and bilateral filter are analyzed and compared.When studying image edge extraction operator,an improved scheme is proposed to overcome the shortcomings of traditional Canny operator and applied to the edge extraction of workpiece.Then,for the target recognition and location problem,the SURF algorithm is used to complete the target recognition work.The location principle of monocular vision and the calibration method of camera are described.Zhang Zhengyou calibration method is used to complete the calibration of the camera,and the internal and external parameters of the camera are obtained.The method of feature matching is used in monocular location.The depth information of the target is calculated by decomposing the essential matrix,and the centroid coordinates of the target are extracted.After that,according to the kinematics principle of industrial robot manipulator,utilizing D-H method to model the robot,and using ADAMS software to carry out kinematics simulation,the established forward kinematics model was verified.A soft competitive ART-RBF learning algorithm is used to solve the inverse kinematics of the manipulator.In order to clearly understand the working range of the robot,the threedimensional workspace of the manipulator is obtained by using MATLAB.Finally,the laboratory equipment is used to build a vision-guided industrial robot system to complete the grasping test of the model.In this paper,a vision-guided industrial robot system is built,and machine vision system and industrial robot system are studied.The rationality of the built system is confirmed by tests,which supplies a reference for the design of industrial robot system based upon vision in the real manufacture.更多还原