A stereo-vision system to automate the manufacture of a semitrailer chassis

This paper describes the optical measuring system that has been developed to measure the position of a mobile drilling unit used during a semitrailer chassis manufacturing process. This measurement system uses optical stereometric techniques to determine, through triangulation, the mobile drilling unit position from images captured by two cameras that incorporate the system. These two cameras have a swivel degree of freedom that allows them to follow continuously the mobile drilling unit. To reduce costs and simplify system design, rather than use two encoders to measure the rotation angles of the cameras, fixed reflective markers have been used. The three-dimensional coordinates of these reflective markers are constant and known; this way, it is possible to determine the cameras reference system orientation anytime from the image coordinates of the markers projections in each camera images. The measurement system has been tested using a prototype which has the same characteristics than the tools that will be used during the manufacture of the semitrailer chassis.     

阵地地貌反求测量中的相机标定

针对阵地地貌测量重构中的相机参数标定问题,基于相机的内参数模型提出了一种多相机并联阵地地貌测量标定方法.该方法基于主动视觉原理,利用测量支架在线性独立位置安装4部相机,并采用并联装置驱动相机同时拍摄,从而实现与三次独立线性平移效果相同的拍摄.根据该标定方法设计了阵地地貌测量标定实验,实验结果表明两个方向尺度因子比值的相对误差＜3.87%,主点坐标绝对误差在4个像素的范围内线性,畸变因子的变化范围在3倍以内,该方法方便、快捷,适合于阵地地貌测量的标定.     

长度约束下双相机外参数不稳定定向

动态立体视觉测量技术在几何量测量特别是大视场测量中应用越来越广。在分析动态双相机立体视觉测量相对定向算法的基础上,提出了一种基于一定数量距离约束的相机不稳定定向算法。在测量场中固定放置一定数量的一维标定靶标,双相机获取标定靶标上点的中心坐标,由极线几何约束快速准确解算双相机外参数,每次测量根据标定靶重新定向相机,由靶标补偿相机在测量过程中的位置变动,保证在测量场中即使相机站位发生位置变化仍可精确测量。实验中对分布范围2.5 m×1.6 m×4.5 m测量场中90个自反射目标点进行测量,选用5对点间距离作为距离约束,并与V-STARS/S进行结果比对。结果表明在相机站位发生明显变动情况下,90个空间点三维重建位置误差小于0.160 mm,标准差小于0.042 mm。将该方法应用于合成孔径雷达测量,得到相近结果。     

Using photodetectors in Shack-Hartmann wavefront sensors

High-resolution cameras are used as photodetectors in the image recording plane of a Shack-Hartmann wavefront sensor to record centroid coordinates which provide basic information for wavefront reconstruction. A comparative analysis is made of the accuracy of determining the centroid coordinates for CCD and CMOS cameras. The modes of instantaneous sampling and frame-by-frame accumulation of information from the cameras are considered.     

阵地地貌反求测量中像机标定技术研究

针对阵地地貌测量重构中相机参数标定问题,基于像机的内参数模型,提出了一种多相机并联阵地地貌测量标定方法。该方法基于主动视觉原理,利用测量支架在线性独立位置安装四部相机,并采用并联装置驱动相机同时拍摄,从而实现与三次独立线性平移效果相同的拍摄。根据该标定方法设计了阵地地貌测量标定实验,实验结果表明fu /fv的相对误差小于3.87%,u0、v0绝对误差在4个像素的范围内,s的变化范围在3倍以内,该方法适合于阵地地貌测量的方便、快捷标定。     

摄影测量的动态测量应用

摄影测量技术是通过空间交会的原理计算出空间点的三维坐标.借助于高精度的照相机和快速图像处理器,摄影测量可以实现对多点的实时测量监控,从而计算出被测物的姿态、位置、形状以及它们的变化.通过数学建模,分析了摄影测量的动态测量原理,包括仪器部件坐标系的建立,活动部件坐标系的动态跟踪.并以比利时KRYPTON产品为例,介绍了典型摄影测量系统的应用技术,以及它们在航天航空汽车制造等领域中的应用实例.     

大视场多像机视频测量系统的全局标定

针对大视场多像机视频测量系统的全局标定,提出并实现了一种基于近景摄影测量技术的高精度像机标定方法.首先,设计一种带有标志点的十字架形标定参考物,十字架正反两个表面粘贴有环形的编码标志点和圆形的非编码标志点.然后,在现场用工业摄影测量系统XJTUDP重建出双面标定十字架的精确结构尺寸数据;将标定十字架在公共视场范围内依次摆出多个姿态,并控制多像机同步拍摄各姿态的图像.最后,为大视场像机选择10参数非线性成像模型,对采集的图像组依次进行标定运算,标定出各像机的内参数和畸变参数,并全局定向出各个像机在统一坐标系下的外方位参数.实验结果表明,该标定方法的重投影误差小于0.05 pixel;利用标定结果测量高精度标准尺长度,相对误差小于1/4 000,可满足大视场多像机视频测量对精度和效率的要求.     

基于线结构光参考平面的多摄像机标定方法研究

多摄像机视觉测量系统中不同视觉传感器空间分布广,无公共视角,现场标定十分困难。针对多摄像机标定问题,研究了一种基于线结构光参考平面的灵活标定方法。标定过程中,以空间中同时覆盖相邻摄像机视角的结构光平面作为标定参考基准,在不同摄像机视角中,自由移动平面靶标多次,确保每次移动后靶标与结构光相交,并能在各自摄像机中清晰成像,摄像机拍摄靶标图像。提取靶标图像中角点坐标、激光光条特征点坐标。借助靶标平面与摄像机坐标系外部参数矩阵,求解激光光条特征点在对应摄像机坐标系中坐标。通过结构光基准平面内,不同摄像机坐标系中至少3组非共线特征点坐标信息,求解相邻摄像机外部参数。分别进行标定试验和精度验证试验,试验结果表明该方法切实可行。     

基于单目摄像机的光学立体标定技术

为了构建一种基于单摄像机的三维视觉系统,提出了重构光学标定点三维坐标、计算重构误差和优化标定参数的方法。通过移动高精度位移平台,构建三维视觉系统,计算左右摄像机位的初始参数,引入质心坐标法计算两摄像机位间的旋转平移矩阵,利用最优三角剖分法重构光学标定点的三维坐标,计算并最小化重构误差,对标定参数进行优化。实验表明：重构误差直接反映了三维重建的效率,该方法的精度和鲁棒性得到了显著的提高;实物外形测量的误差从0.1123 mm减小到0.0191 mm,标准差从0.1838减小到0.1275。该方法适合应用于三维视觉系统标定质量的评估。     

桥式起重机的近景摄影测量方法研究

与常见的桥式起重机变形检测方法相比,近景摄影测量法具有非接触性、瞬间捕获信息量大且数字图像易存储等优点。文中将近景摄影测量与桥式起重机变形检测相结合,通过DLT(直接线性变换)标定非量测相机,采用序贯分析求解其内方位元素;在桥式起重机上合理布置反光标志点,并使用非量测相机获取数字立体像对;通过软件获得数字图像上各个标志点的坐标仪坐标,采用光束平差法求解标志点的物方坐标,构建桥式起重机三维模型,计算主梁变形量。将近景摄影测量法与水准仪、全站仪方法进行对比,证明基于近景摄影测量的桥式起重机变形检测方法能够满足桥式起重机精度要求,为桥式起重机变形检测提供了一种高效的方法。     

NAO的单目视觉空间目标定位方法研究

为了实现单摄像头工作的双足机器人NAO的视觉定位,采用单目视觉技术,建立一个空间点定位模型。根据摄像机的小孔透视模型,将图像中的二维坐标通过几何关系映射为机器人坐标系中的三维坐标,实现NAO基于单目视觉技术对空间目标定位。进行了定位实验,实验结果误差在允许范围之内,验证了该定位方法的实际可行性。     

GPS双目摄像机标定及空间坐标重建

为减轻双目摄像机标定过程中对高精度靶标的依赖,实现摄像机参数的精确标定,并对空间坐标进行高精度重建,提出一种GPS双目摄像机标定及空间坐标重建方法,采用GPS代替2D或3D靶标进行双目摄像机标定。将GPS的位置在视场中任意移动,由被标摄像机拍摄多组含有GPS的图像,利用空间三维坐标与图像二维坐标间的映射关系,结合摄像机成像模型和双目摄像机标定原理,标定出双目摄像机参数,并对空间坐标进行精确重建。通过空间重建坐标与GPS实际测量值之间的相对距离误差,对重建精度进行检验。实验证明,该方法能够克服双目摄像机标定过程中对高精度靶标的依赖,空间重建坐标具有较高的精度,相对距离误差从1.56%减小到0.52%。     

A flexible technique for calibrating relative position and orientation of two cameras with no-overlapping FOV

We propose a flexible new method to calibrate the relative position and orientation relationship of two cameras with no-overlapping field of view (FOV), which are called relative pose parameters in this paper. The objective of this method is to unify the local coordinate frames of the two cameras to be under one global coordinate frame. Additionally we design a flexible target composed of two short 1D bars with equally placed light spots and one long linking pole to assist the calibration. The target is adjustable according to the actual position and orientation of the two cameras, based on the principle that the two 1D bars can be shot clearly at the same time by their corresponding cameras, which will largely increase the calibration accuracy. The proposed approach consists of two steps. Firstly we use the invariant constraints on the angle and distance between the two short 1D bars to give a rough estimation on the relative pose parameters of the two cameras, then we use more restrictions on the invariant distances between light spots on the two 1D bars respectively to refine the estimation on both the relative pose parameters and the structure parameter of the targets. This technique offers a significant flexibility and tight space advantage compared to other calibration methods applying extremely large 2D or 3D targets for wide view calibration, which are inefficient in some conditions. And equally important, this approach still obtains acceptable calibration accuracy. Result of the experimental calibration shows that the system calibrated by this method reaches an accuracy of 0.108 mm when measuring a distance section of 1040 mm.     

离轴三反时间延迟积分CCD相机内方位元素和畸变的标定

由于测绘相机的关键几何参数内方位元素和畸变的标定精度决定相机的立体测绘精度,本文提出了一种离轴三反时间延迟积分( TDI) CCD相机内方位元素和畸变的标定方法.介绍了离轴三反TDICCD相机的光学系统和像面拼接方法,明确了该相机内方位元素和畸变的含义.建立了标定系统及相应的数学模型,应用最小二乘回归法求得了内方位元素和畸变的表达式.利用提出的方法标定了相机的内方位元素和畸变,并对标定误差进行了分析.结果表明:该方法对主点的标定精度可达1.0 μm(1σ),对主距的标定精度可达2.0 μm(1σ),对畸变的标定精度为2.3 μm(1σ).结果显示提出的标定方法快捷且有效.     

Calibration and analysis of eccentric error of the laser rotary-scanning measurement system

A laser rotary-scanning measurement system was developed for the reverse engineering of 360° objects. The system is constructed by an optical head and a rotary indexing. The optical head is composed of a laser diode strip-light projector and dual CCD cameras. Based on the principle of structured-light triangulation, a laser line is projected onto the object upon which the distorted line is captured by dual CCD cameras from left and right simultaneously. By processing a series of line fittings from the discrete angular positions of an object, the entire 3D profile can be reconstructed. Since the actual space coordinates of the object are computed according to the geometric relationship between the coordinate of optical head system and the coordinate of the rotary indexing systems, if these two coordinate systems are not in good alignment, errors in the computed coordinates will be introduced.This paper describes the influences of the alignment and eccentricity errors of the laser rotary-scanning measurement system on the computed geometrical profile. Calibration procedures are then proposed to adjust the alignment to avoid image distortions and thus enhance the system accuracy. Experimental results show that this easy-to-use calibration procedure can significantly improve the accuracy of the system.     

Wheel alignment inspection by 3D point cloud monitoring

Today’s wheel alignment inspection systems adopt various computer vision technologies. They, however, require high-end cameras, precisely manufactured targets, and massive calculation loops because they rely on low-dimensional data (two-dimensional images) for measuring higher-dimensional information (three-dimensional orientation) of the wheel posture. To improve this, a simple and inexpensive method using a consumer-grade depth-sensing camera such as Kinect is presented. It directly utilizes point clouds generated from its range image stream. All points within the region of interest (ROI) contain geometrical information of the wheel and are used for the alignment inspection procedures. Its feasibility is evaluated by examining whether the orientation could be aligned to the desired orientation using only the point cloud data. For verification, a one-wheel-based prototype was implemented, and comparative experiments with an existing commercial system were conducted. The experimental results showed that the proposed method provides satisfactory performance. We believe that the proposed method is feasible for practical usage and has a great potential to be an effective alternative to existing wheel alignment inspection methods.     

基于射影几何的目标点三维坐标值测量

针对机械手抓取的目标物其空间位置难以定位的现状提出如下方法：在由双摄像机采集到的两幅图像中利用模板匹配找到目标物,测定其中心点的图像坐标,将其坐标值代入摄像机的成像模型中进而算出该点的三维坐标值。该方法具有精度高、适用性强的特点。实验结果表明,用该方法确定的目标物位置能够达到机械手抓取所需的精度要求。     

一种基于位姿反馈的工业机器人定位补偿方法

为了提高工业机器人的绝对定位精度,提出了一种基于末端位姿闭环反馈的机器人精度补偿方法。该方法通过激光跟踪仪测量实时跟踪机器人末端靶标点的位置来监测机器人末端的位姿,并通过对靶标点的实际位置和理论位置进行匹配获得机器人末端的位姿偏差。工业机器人系统与激光跟踪测量系统通过局域网进行数据通信,并根据位姿偏差数据对机器人末端的位姿进行修正。最后通过实验对基于末端位姿闭环反馈的机器人精度补偿方法进行验证,实验表明,经过位姿闭环反馈补偿后机器人末端位置误差最大幅度可以降低到0.05mm,姿态误差最大幅度可以降低到0.012°。     

ENT 手术机器人视觉系统研究

设计一套耳鼻喉手术机器人视觉系统,采用参数约束化方法对双目CCD系统进行标定,提高系统精度。利用光心点到投影线距离之和最小的约束方程,缩小重构范围,提高三维空间坐标的求解精度。提出在喉镜口端面建立欧氏空间坐标的方法。将视觉系统与机械人控制系统进行了集成。     

基于立体视觉的六自由度平台位姿检测基础研究

位姿是六自由度平台的一项重要性能指标,其全方位、全行程的动态精确检测,对于六自由度平台的精确控制及广泛应用具有十分重要的意义,介绍六自由度平台,提出并着重讨论采用立体视觉对六自由度平台位姿进行检测的原理及具体步骤。实验结果表明,对于六自由度平台位姿的检测。立体视觉是一种可靠、快速、简单、可行的方法。