【作者】 陈菲；

【导师】 杨兆军；

【作者基本信息】 吉林大学 ， 机械制造及其自动化， 2009， 博士

【摘要】 本文结合国家高技术研究发展计划(863计划)“大型曲面自主研抛作业微小机器人技术”,以实现微小机器人在大型曲面上的自主定位为研究目标,创新性的将计算机视觉坐标测量技术应用于移动机器人空间定位方法中,提出了基于六个特征点平面距离约束的微小研抛机器人单目视觉空间定位新方法。实验证明该方法能够满足微小机器人在大型曲面上的定位要求,定位算法具有较好的鲁棒性,也可为多个移动机器人系统同步定位提供一种新的解决思路。论文根据PnP(Perspective n Points)理论,以摄像机中心投影模型为基础建立了视觉定位系统的数学模型,提出了基于六个平面特征点距离约束的参数建模方法和机器人空间位置坐标及姿态的求解算法。结合微小研抛机器人空间定位的实际需要,建立了摄像机标定的内外参数模型,提出了基于非线性模型的摄像机平面网格标定新方法。通过大量的摄像机标定实验,对标定方法的精度进行了研究,实现了在整个工作空间内对摄像机内外参数的校准。针对定位系统采集的灰度图像特点,提出具有较强抗干扰性、基于连通成分标记的自适应阈值图像分割方法,在实现图像分割的同时,快速、准确的将六个特征像点区域区分开来。对圆形和椭圆形目标的亚像素定位方法进行了研究,采用双线性插值灰度平方加权质心算法,对成像点的中心坐标进行了亚像素求解,仿真分析和实验结果均验证了该方法能够有效降低计算结果的离散程度,实现特征点的1/10倍像素细分,满足了机器人定位系统对特征点成像中心精确定位的要求。在大型曲面上进行了综合定位实验,并对定位系统误差产生原因进行了系统的理论分析,尤其针对当辅助定位特征点的光轴与摄像机光轴成一定角度时、亚像素定位算法计算出的特征像点的中心坐标与实际发光点成像中心坐标间存在较大偏差这一现象,提出了角度偏置误差补偿理论和预估-即时修正补偿方法,通过实验手段建立了误差补偿数学模型,实验证明了该方法能够有效提高定位系统的精度和系统的稳定性。课题研究工作得到了国家高技术研究发展计划“大型曲面自主研抛作业微小机器人技术”(项目编号:2006AA04Z214)和国家自然科学基金项目“自定位微小研抛机器人精整加工大型曲面研究”(项目编号:50575092)的资助。更多还原

【Abstract】 With the development of science technology, the application fields of robot have been enlarged constantly. For large free-form surfaces, the polishing process is limited to range of the processing equipment, so it is very important to develop robotic polishing process technology of free-form surfaces in order to realize the automatic polishing proess of large free-form surface. In order to realize free movement of polishg robot in the large free-form surface, autosensing and self-positioning, the spacial localization methods of polishing robot is a preprquisite to realize autonomous navigation and automatic control of polishing robot in a certain working space. There is the significance of theory and reality in space localization fields of moving robot. Localization is an important content in the research field of robotics and it also is premise and guarantee of navigation and path planning. It is the real-time and high precision localization of mobile robot which make the mobile robot mounting industrial robot accurate working in large range possible. In addition, the perfect combination of the mobile robot and the industrial robot make have broad perspectives in manufacturing, unmanned factory and spatial working application. But the navigation systems and localization technology are a difficult problem both the individual mobile robot and muti-robot system.This paper applys the computer visual coordinate measuring technology to the spatial localization method of automated mobile robot creatively in order to improve the autonomous ability of robot localization and roboustness of localization algorithm, while to reduce the effect of nonsystematic error to the localization accuracy and offer an effective synchronous localization method of muti-robot system. The spatial localization method of micro-polishing robot based on constraint of characteristic point distance is researched n this paper with visiual coordinate measuring technology in order to improve localization precision and reat-time of localization system. The localization problem of micro-polishing robot is researched profoundly in this paper. The work of the paper mainly involves several aspects:Firstly, the domestic and overseas present status and development trend of visual localization method of mobile robot is analysed in this paper. A novel localization method of micro-polishing robot combined with the research subject, which is restricted within certain working space, is presented in this paper. On the basis of pinhole camera model, a new mathematical model of vision-localization of automated polishing robot is established. The vision-localization is based on the distance-constraints of feature points. The method to solve the mathematical model is discussed.According to the perspective principle of the camera, the mathematic model of localization system is established and solutions are solved by the mathematic method. The theoretical background of mathematic model is discussed; the problem of Perspective n Points is analyzed in the statement of problem, two defination methods and the present status respectively. The arrangement of six optical characteristic points in the micro-polishing robot are discussend secondly, the feasiblilty and reliability of the arrangement method are demonstrated both in theory and in experiments. The mathematic model of localization system is established according to the definantion of PnP and the corresponding center coordinate of image points. The optimization soloutions based on the total Least Squares are derived by the Singular Value Decomposition and the position and the posture of robot in working space are determined lastly.The calibration accuracy of camera parameters is the key part of high accuracy optical measurement On the basis of the analysis of various calibration methods, the plane net calibration method of nonlinear model including Radial Lens distortion and Tangle Lens Distortion is presented after the establishment of mathematical model, and the concrete resolved method is reduced. The following experiments show that the calibration method is high precision and can meet the technology requirement of robot localization.The presented localization method is based on the distance constraint theory of characteristic points. The image shape on the image coordinate is approximate to ellipse. On the basis of the present localization methods of sub-pixel center coordinate, the adaptive threshold image segmentation method and the gray square weighted algorithm of bilinear interpolation are presented in this paper. According to the given position of six optical points, the position matching algorithm is discussed .Canny operator rapid localization method and the principle of sub-pixel edge detecting algorithm based on Zernike orthogonal moments are presented in this paper. The center orientation method of elliptical curves fitting based on the Canny-Zernike algorithm is given which is closed to circular markers. Finally, the validity and precision of elliptical curves fitting algorithm based on Canny-Zernike operator is studied with computer simulation analysis and experimental research, the sub-pixel edge localization results are compared to gray weighted algorithm. The results indicate that the center localization precision of circular marker based on Canny-Zernike orthogonal moments edge detection operator is much better than traditional edge detection operator and the position precision of the operator is better than 0.05 pixels. The experiments show that the gray square weighted algorithm of bilinear interpolation can meet the requirement of accuracy localization of image points center.In order to improve the accuracy of the system, this paper analyzed the error factors of the measurement system, researched the systematic error caused by feature points’angular-displacement, and proposed the error compensation method. Aim at the inherent defect of single vision system which is the error of Z axis are large, the reson is analyzed both in quantatitive analysis and qualitative analysis. The method of super plane fitting is present in order to improve the measuring accuracy in z direction. At first, based on the LED construction and imaging principle, it analyzed refraction and reflection factors. Then established the refraction compensation function by theoretical analysis and fitted the reflection compensation function by experimental tests. After that, it researched the comprehensive compensation function depended on the two functions above. At last, on basis of the predictor-corrector method, this paper got the probe center coordinate through real-time modifying feature points’angular-displacement-error, and verified the feasibility of this method through the accuracy measurement experiments.Finally, the resolution of image system, the geometrical distortion of projected image, the deviation of optical characteristic points position, the center coordinante error of characteristic points and the quantification error are analyzed systematically.Finally, the total experimental project plan and procedure is discussed. The experimental results show that the presented model and solution method is correctness and high precision. The method is also suitable for localization of mobile robot in workspace. Many error sources exist in measuring system, so establishing rational error model and improving location precision of points are the future targets of research.In this paper, the main innovations are as follows:1. This paper applys the computer visual coordinate measuring technology to the spatial localization method of automated mobile robot creatively in order to improve the autonomous ability of robot localization and roboustness of localization algorithm, while to reduce the effect of nonsystematic error to the localization accuracy and offer an effective synchronous localization method of muti-robot system .2. The mathematical model of mobile robot spacial localization based on six optical characteristic points is presented in this paper and the optimization soloutions based on the total Least Squares are derived by the Singular Value Decomposition and the position and the posture of robot in working space are determined. The total experiments of localization system show that the method can improve the autonomous ability of robot localization and real-time of localization.3. The error compensation method of super plane fitting is presented in order to improve the measuring accuracy in z direction. The method advances the stability and the veracity of the center coordinate of characteristic image point and the measuring accuracy of localization system can be improved.4. The angle between the camera optical axis and the optical axis of the characteristic point changes gradually. The error between the real center coordinate and the calculative center coordinate of the characteristic image point is caused because the inner wall reflection of the optical characteristic points exists.The things account for its occurrence are analyzed, the error compensation theory is researched and the mathematical model of error compensation are established. The experiments show that the method can increase the accuracy of the localization system.更多还原