正交柱面成像式动态坐标测量方法研究

发布时间：2018-04-01 18:29

本文选题：正交柱面成像　切入点：动态坐标测量　出处：《天津大学》2016年博士论文

【摘要】：推动工业由加工制造向智能制造的转型升级是未来工业的发展趋势之一,作为智能制造的典型应用,智能化装配和运动过程监控要求实现对多个目标点的同步、实时测量。现有的坐标测量方法,如:激光跟踪全站式测量、光电扫描网络式测量、立体视觉测量,在对运动物体进行测量时,分别面临着单目标、非同步测量、测量频率低的局限性。所以必须探索一种高性能动态坐标测量新方法,以满足不断发展的工业需求。本文针对现有动态坐标测量方法的局限性,从传感层面出发,对视觉测量中面阵图像传感器二维测角功能进行分解,采用两个正交放置的线阵CCD(Charge Coupled Device)结合正交柱面成像光学系统,实现目标点的二维角度测量,多角度交会实现三维坐标测量。新方法从原理上保证了多目标高频、同步测量的可行性。以正交柱面成像式测量系统为研究对象,在对其误差分析的基础上,依次从相机校准、动态性能提升、多目标跟踪三个方面展开研究,详细阐明了各自的主要问题、解决办法和具体实现,建立了较为完整的动态坐标测量方法体系,并通过仿真和实测实验进行验证,实验结果表明,正交柱面成像式测量系统具有良好的测量精度和动态性能,能够完成对多目标的实时跟踪。论文主要研究内容如下:1、以所提出的正交柱面成像式测量系统为研究对象,对其角度交会坐标测量原理进行研究,建立了角度测量模型,描述了坐标重构方案,并从应用层面给出系统实现。2、分析了测量系统的主要误差源,对单相机的误差因素进行分析以便寻求更加准确的相机校准方法,对多相机布局进行分析以便研究坐标误差传递关系,两者融合提高系统测量精度。3、针对传统相机校准方法在进行复杂镜头畸变校正时所面临的局限性,从准确复现成像系统投影映射这一相机校准的本质出发,提出了基于径向基函数映射的相机校准方法,为测量系统提供了精度保障。4、考虑到动态测量中的随机误差,引入卡尔曼滤波算法对系统动态性能进行提升,研究了基于线性卡尔曼滤波和扩展卡尔曼滤波的动态坐标测量方法。5、为解决柱面成像系统普遍面临的多目标识别问题,提出了基于扩展卡尔曼预测的多目标识别跟踪方法,以实现多目标物点像点之间的匹配识别,针对测量过程中可能出现的光斑交叉和粗大误差情况,给出了相应的解决办法。
[Abstract]:As a typical application of intelligent manufacturing, intelligent assembly and motion process monitoring require synchronous and real-time measurement of multiple target points.The existing coordinate measurement methods, such as laser tracking full-station measurement, photoelectric scanning network measurement, stereo vision measurement, are faced with the limitations of single target, asynchronous measurement and low measurement frequency respectively.Therefore, a new method of dynamic coordinate measurement with high performance must be explored to meet the continuous development of industrial demand.Aiming at the limitation of the existing dynamic coordinate measurement methods, this paper decomposes the two-dimensional angle measurement function of the plane array image sensor in visual measurement from the sensing level.Two orthogonal array CCD(Charge Coupled devices and orthogonal cylindrical imaging optical system are used to realize the two-dimensional angle measurement of the target point and the 3D coordinate measurement by multi-angle rendezvous.The new method ensures the feasibility of multi-target high frequency and synchronous measurement in principle.Based on the error analysis of orthogonal cylindrical imaging measurement system, this paper studies camera calibration, dynamic performance improvement and multi-target tracking in turn, and explains their main problems in detail.A complete system of dynamic coordinate measurement method is established and verified by simulation and practical experiments. The experimental results show that the orthogonal cylindrical imaging measurement system has good measuring accuracy and dynamic performance.Can complete the multi-target real-time tracking.The main contents of this paper are as follows: 1. Taking the orthogonal cylindrical imaging measurement system as the research object, the principle of the angular intersection coordinate measurement is studied, the angle measurement model is established, and the coordinate reconstruction scheme is described.The main error sources of the measurement system are analyzed, and the error factors of the single camera are analyzed in order to find a more accurate calibration method for the camera.This paper analyzes the layout of multi-camera in order to study the coordinate error transfer relation, and improves the measuring accuracy of the system by combining the two methods. Aiming at the limitation of the traditional camera calibration method in the correction of complex lens distortion,Based on the essence of camera calibration, a camera calibration method based on radial basis function mapping (RBF) is proposed, which provides a precision guarantee of measuring system .4, considering the random error in dynamic measurement.Kalman filtering algorithm is introduced to improve the dynamic performance of the system. The dynamic coordinate measurement method based on linear Kalman filter and extended Kalman filter is studied. In order to solve the problem of multi-target recognition in cylindrical imaging system, the dynamic coordinate measurement method based on linear Kalman filter and extended Kalman filter is studied.A multi-target recognition and tracking method based on extended Kalman prediction is proposed to realize the matching and recognition of multi-object image points. The corresponding solutions are given for the possible cross-section and gross errors of light spots in the measurement process.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】