【作者】 王芳荣；

【导师】 赵丁选；

【作者基本信息】 吉林大学 ， 机械电子工程， 2007， 博士

【摘要】 本文结合国家自然科学基金项目“遥操纵6自由度液压并联机械手的力觉双向伺服控制”,对机器人视觉三维测量、空间定位、尺寸测量等重要技术进行理论和实验研究,设计了基于结构光的单CCD三维轮廓测量系统,通过对测试系统的标定,得到了被测物的畸变条纹和被测物的三维形貌的对应关系,并根据解析几何知识和实验硬件系统特性,提出了一种对旋转平台系统进行坐标校正的激光刀口斜率拟合校正法,并对几种不同材质的物体进行了三维重构,重构结果表明,该方法对于不同反射率的机械件的三维重构,均有良好效果;设计构建了基于双目立体视觉的空间定位系统,提出了一种将空间点位置与屏幕坐标相关联的数学模型,并依据灰度矩法,实现了激光光斑形心的精确定位,修正了光斑形状变化的影响,定位精度达到2%,稳定性高;设计了一种激光衍射法检测细丝直径的系统,提出了一套测量细微孔径尺寸的数据测量处理方法,将质心法应用于条状衍射斑处理,使细丝尺寸检测精度,摆脱了CCD像元尺寸的局限,达到了亚像素水平,实验表明,该方法具有很高的测量精度(小于1%),测试速度也很快,可以用于自动实时的在线测量,具有很好的实用价值和应用前景。更多还原

【Abstract】 In recent years, the robot is playing an increasingly important role in industry application. However, more than 90% of the robots is used to fix workpiece on the precise position, rely on the precisely getting and laying workpiece, efficiency and product quality can be improved. These robots do not have the ability to get the information from the outside world. Because of the changes in working conditions, the production process interruption or failure leads to heavy losses. So the robots need to have a higher ability to identify targets and to accept the information from the outside word as a human, and they need to have autonomy. So people pay more attention to intelligent robot.Human gets more than 75% information by visual sense, complex tasks can be completed under the lead of human vision. Human hopes that the machine can have the ability of recognition and perception to do dangerous work in stead of human. intelligent robot is applied in many fields such as robot remote operation、removal radioactive nuclear wastes、space exploration、deep-sea exploration、and the battlefield rescue、fire rescue、explosives remove、medical and health work、vehicles experiment and so on. So the exploitation of visual intelligent robots is very important. As an important component of the robot, visual system can get object information, including its shape, position, posture, movement, and it also can describe、save、identify and understand them. This paper is based on the National Natural Science Fund project (ID : 50475011) to research and develop visual intelligent robot. On the basis of in a lot of literature, robot 3D vision measurement、spatialization、size measurement is studied, The main contents of this paper are as follows:(1)Design 3D profile measurement system using rotary platform based on the structured light; structure hardware system and found mathematical model; give laser incision slope fitting correction method to correct rotating platform system coordinates. Through experiments collect and process 3D object data, analyze the measurement error. Using reverse engineering method to rebuild small metal products、plastic products、stone、translucent glass. Reconstruction results show that the method have good results for different reflectivity of mechanical parts.(2)Use camera system, design spatialization system based on binocular stereo vision。A spot space orientation system by using double area CCD was introduced. The mathematical model was stablished. Using center of gravity method ,the system make the precise orientation of la ser spot come true .Through the mea sure to different rectangle ,circle and triangle ,the precision of center orientation can be up to 2 % ,and can’t be influenced by different forms of spots. The precision has good steadiness.(3)Combining the Babinet complementary and Fraunhofer slit diffraction theory with the above-mentioned calibration scheme, we create the measurement system of micro-size. Using the system we measure the filament diameter, at first, picture the diffraction fringe, then process the image data with centroidal and curve fitting methods. The method can get the measurement precision.(4)Study the camera calibration method, and compare several different calibration methods; introduce the process of the standard c hessboard free plane calibration, calibrate the test system of the camera, and obtain cameras parameters, and found the relationship between distortion fringes and 3D pattern. Using this model optimize calibration parameters. Experiments have shown that the demarcation process is simple and easy to operate, and can improve the accuracy.(5) Combine laser triangulation measurement and three-dimensional imaging mathematical model, discuss the process from 3D contour information to two-dimensional image information, and determine the characteristic region of two-dimensional image data, and use image processing technology to select、filter characteristic region, and obtain the 3D information of section plane.(6) Use DirectShow technology for data collection and processing, and collect and save visual images; use Matlab and Visual C++ to analyze and process the data.(7) Measure the standard size and surface area of cylinder; analyze system measurement error.In short, this paper describes computer vision 3D measurement、space positioning、size measurement method. The study results have some reference value for visual intelligent robot.更多还原