全场视觉自扫描测量系统

为了适应对大型工件的快速精确测量,提出一种由CCD摄像机、振镜和激光线投射器组成的全场视觉扫描测量系统,通过振镜的转动改变线结构光光平面角度来实现对摄像机视场内被测物体的自扫描测量。利用平面网格靶标和基于交比不变原理的径向排列约束(RAC)两步方法实现对摄像机的标定,根据振镜的旋转角度以及光平面在靶面上的交线确定光平面的方位,从而根据摄像机模型和CCD二维图像可获得光平面上的三维数据。试验结果证明,这种方法具有较高的测量精度,能较清晰地测量大型工件上的主要特征,同时具有便携性,便于生产现场使用。     

线结构光测头的误差补偿

提出了一种线结构光测头误差检测和补偿方法。通过对线结构光测头测量精度的误差来源进行系统分析,确定测量过程中光平面倾斜和光条宽度的变化是产生测量误差的主要因素。提出了倾斜误差的检测方法并设计了检测装置,利用改进的BP神经网络模型实现了倾斜误差的补偿。实验证明,该方法能在很大程度上提高线结构光测头测量数据的精度,补偿后的误差接近光平面倾斜角度为零时的误差。     

基于机器视觉的零件圆形特征尺寸测量系统设计

针对零件尺寸传统测量方法存在测量效率低、测量结果易受人为因素干扰等问题,设计一种基于机器视觉的零件圆形特征尺寸测量系统。该系统在Visual Studio 2017环境下通过C#中Winform窗体设计模块设计测量系统的人机交互界面并联合视觉算法库Halcon,利用C#调用Halcon中相关视觉算法完成对待测零件圆形特征部分圆心、半径的测量。经实验验证,该系统测量精度高、效率快,能够实现对零件圆形特征的测量。     

线结构光与CMM集成测量系统一体化标定方法

线结构光三维视觉与三坐标测量机(CMM)测量系统的集成能够实现对复杂零件表面快速高精度的测量。系统的标定是实现集成测量的关键步骤,包括摄像机内外参数标定,光平面结构参数标定和两个测量系统测量坐标的全局统一。提出了线结构光与CMM集成测量系统一体化标定方法,首先采集多幅不同位姿下带有光条纹的2D靶标图像,然后利用CMM接触测头测量得到某一位姿2D靶标角点在CMM坐标系下的世界坐标,利用交比不变原理求取靶标上特征点的三维坐标,进而求取线结构光视觉传感器测量参数和线结构光传感器测量数据到CMM坐标系下的转换矩阵,实现一次数据采集完成集成系统所有测量参数的标定,标定方法简单且精度和效率较高。     

基于多线结构光的弱纹理物体形貌测量方法

大多数现有的主流结构光测量装置在被测物体的表面上投射单条激光,通过移动装置实现物体表面的扫描重建。为了实现物体表面的在线快速重建,将多线结构光和双目立体视觉理论相结合,设计了一种新的三维测量方法。提出了一种结合形态滤波和Zhang-Suen细化算法的光条骨架提取方法和基于光条序列关系的光条匹配算法,并通过极线约束实现光条上特征点的精确匹配。通过实验验证了方法的有效性,相对误差在3%以内。     

基于立体定向靶标的探针式多视场三维视觉测量系统

针对含有遮挡区域、深孔及凹槽等特征的多面体或回转体物体,设计了一套基于立体定向靶标的探针式多视场三维视觉测量系统,并阐述该套测量系统的结构组成和基本工作原理。首先,基于近景摄影测量技术建立立体定向靶标的6个单元模型,计算靶标各侧面角点在各自单元模型内的坐标,再通过单元模型的链接和光束平差,获取全部角点的精确全局坐标,并将其作为立体定向靶标的全局控制点。然后,设计了利用共面角点辅助定位的探针,仍基于近景摄影测量技术解算出角点和测头在探针坐标系中的精确坐标。最后,利用共面的棋盘格角点与其像平面之间的单应性矩阵,推导全局坐标系、探针坐标系各自与相机坐标系的位姿关系,进而求得探针测头的全局坐标。以量块（量棒）的标准长度作为评价指标,在2m×1.5m的视场范围内测量精度优于0.1mm。测量实验表明,多视场三维视觉测量系统用于具有回转体结构特征的水壶测量,能够获取水壶表面全部区域的点云数据。     

一种线结构光自扫描测量系统的研究

为了实现对复杂自由曲面的快速精确测量,提出了一种由CCD摄像机、振镜和激光线投射器组成的全视觉自扫描测量系统。该系统利用光学三角法确定了一种新的测量物体表面三维坐标的方式。基于该方式,此测量系统使用平面靶标实现了对摄像机内部参数的标定,借助振镜转动驱动激光平面完成了对摄像机视场内被测物体的扫描,最终根据激光平面方程和计算机二维平面图像信息获得了被测物体的三维数据。试验结果证明,该系统能以较快速度实现自由曲面的精确测量。     

多高度小零件视觉测量方法设计

设计一种综合应用激光三角法和二维机器视觉测量技术测量方法,实现对多高度小零件二维几何尺寸的测量。首先在测量高度范围内,利用线结构光进行高度测量标定,采用多项式拟合建立线结构光成像位置与高度的关系;用精密圆形靶标对不同高度平面进行二维尺寸标定,分别考虑不同高度平面坐标系映射到图像坐标系后光学放大率不同和坐标系间的不平行,CCD摄像机阵列垂直方向与水平方向不一定垂直的情况,以及光学镜头畸变等得到不同平面所在高度的二维尺寸测量标定参数,由此建立起各二维尺寸标定参数和高度的函数关系。对具有不同高度的小零件进行测量,并与三坐标测量机结果进行比对,多次测量的高度最大误差小于0.032mm,二维尺寸最大误差均小于0.08mm。     

光笔式大视场三维视觉测量系统

针对先进制造业对大型装备大范围精密尺寸测量需求,根据双目立体视觉测量原理设计了一种光笔式大视场三维视觉测量系统.基于透视投影变换下的时针顺序和共线性不变量设计了光笔特征点空间分布模式,实现了特征点的准确识别与接触探头坐标的计算.应用双目立体视觉传感器的透视投影和齐次坐标三维测量模型,以一维基线尺靶标自由移动和基准长度约束为核心,通过本质矩阵E的线性求解结合非线性优化实现了其结构参数的现场精确标定.研制了由光笔、双目立体视觉测量系统、便携式三脚架、一维基线尺靶标和测量软件构成的大视场三维视觉测量系统,完成了机器人本体表面三维数据的稠密测量实验,在7 m×4.7m测量范围内系统的测量精度优于0.2 mm.实验显示设计的光笔式大视场三维视觉测量系统在光笔结构、发光点识别方法和系统标定方法上均具有新的思路.     

基于全局控制网的三维形貌测量方法

由于传感器成像视场范围的局限性,双目视觉传感器在一次成像中往往不能获得大尺寸物体的整体形貌测量结果,鉴于此,研究并实现了一种通过建立全局控制网来实现对大尺寸物体的三维形貌进行测量的方法。实验结果表明：基于全局控制网的测量方法速度快、精度高,间接地扩大了双目视觉传感器的测量范围,并且有效地克服了大尺寸三维形貌测量拼接精度低的缺点。该方法适用于大尺寸物体的三维形貌测量,能够快速、简单、精确地获得被测物体表面特征点的相对位置关系和三维尺寸信息,进而恢复被测物体的三维形貌。     

数字X线影像仪中激光扫描光学系统的设计

数字X线影像仪是计算机和激光技术快速发展的产物,而其中的激光扫描光学系统是其核心技术之一。针对数字X光影像仪的使用特点,对激光扫描系统中的Fθ镜头、光束扩展器、扫描器这几个关键部件进行了较详细的论述分析,提出并解决了其中一些关键问题,如光阑位置浮动对像质的影响、影响系统扫描光点大小的因素、扫描器的确定等。最后用ZEMAX光学设计软件对系统的光学性能进行了设计模拟,得到扫描光斑直径小于0.1mm、焦距和视场满足线性关系的设计结果。像质评价分析结果表明,所设计的镜头像质优良,轴上与轴外质量相当,像质达到衍射极限。     

光学三维测量技术与应用

光学三维测量技术是一种可视化的测量技术,能完成复杂形体的点、面、形的三维测量,能实现非接触测量,在制造业和航空航天领域得到广泛应用。航空三维激光扫描与摄影测量技术利用激光测距原理和航空摄影测量原理,与基于全球定位系统(GPS)和惯性测量装置(IMU)的机载定位定向系统(POS)联接,可以快速获取大面积地球表面三维数据。本文对光学三维测量技术和航空三维激光扫描技术做一简单介绍,给出三维激光扫描技术在航空航天领域的几个应用。     

Measurement and control technology of the size for large hot forgings

Aiming at the different characteristics and measurement requirements of large hot forgings, a size measurement and control method is proposed. For irregular forgings, a measurement model based on line laser scanning is first constructed and then a solution of the optical center motion trail on the basis of the chessboard reference plane is proposed. The motion trail is solved according to the feature point coordinates on the reference plane in different frames. The overall three-dimensional size is finally achieved by combining the measurement model with the optical center motion trail. In addition, for regular forgings, under the guidance of images obtained by CCD, the two-dimensional size can be achieved according to the displacement of the line laser along the rail. Moreover, the size control of the forgings can be realized by rolling reduction control and the working position location. The measuring method proposed is feasible according to the experimental results.     

Analysis of the geometrical features detectability constraints for laser-scanner sensor planning

This paper presents a planning approach for automatic synthesis of the viewpoint locations of a laser scanner that satisfy the detectability constraints of the geometrical features of a given object. Dominant constraints taken into consideration were the laser scanner field of view, depth of view, and the maximum permissible view angle. The approach is task driven and sensor model based. Information is given about the scanned object and its selected features, as well as about the optical characteristics of the sensor. Planning is presented as an optimisation problem where the optimal laser-scanner viewpoint must satisfy all the constraints simultaneously. The approach is demonstrated by scanning a 2D feature, and through planning an experiment in which robotic assembly of a car door is assisted by a laser scanner.     

Path planning of multi-patched freeform surfaces for laser scanning

In this paper, a new method for scanning a complex surface model with multiple patches is proposed. In scanning of freeform-shaped models, it is difficult to generate a measuring plan by human intuition due to the complexity of the part. The developed method calculates the number of scans, scan direction, and scan path for a stripe-type laser scanner. The user can designate the surface patches that need to be scanned. In generating the scan plan, surface information is first extracted from the CAD model automatically. Other information such as, design knowledge, functionality and machine parameters are provided interactively. Based on this information, the original surface model is sampled in the u, v parameter domain and the sampled points are grouped using the view angle of the laser scanner. The scan direction is determined by calculating the global mean vector of the sampled points that belong to a group. The range of the view angle can be varied to enhance the quality of captured point data. The optical scan paths for each scan direction are calculated by projecting the sampled points onto a 2D plane. The optical constraints are also verified to guarantee the generated scan plan.     

激光扫描追踪人体目标位姿的算法研究

针对人体目标位姿的追踪问题,对二维激光扫描仪的工作原理进行了归纳,对所测取数据的分析算法进行了研究,设计了一种基于激光图像扫描的新型算法,利用C语言与Matlab混合编程,通过对人体目标肩部截面进行椭圆拟合,实时测取了运动人体目标的距离、角度和方位信息,并通过Matlab GUI显示.二维激光扫描仪通过旋转的光学部件发射光脉冲,形成了二维的扫描面,以实现区域扫描及轮廓测量的功能.研究结果表明,利用激光扫描仪测取的人体目标的位置信息与真实的人体目标的位置信息相比基本吻合,距离范围为0m～4m,距离精度优于0.05 m,角度范围为240°,角度跟踪精度优于2°,方位范围为360°,方位精度最高1 °,该算法可用于需要对人体目标进行实时追踪的工程应用中,例如用于对人体目标实时跟随的移动平台或机器人中.     

A multi-view time-domain non-contact diffuse optical tomography scanner with dual wavelength detection for intrinsic and fluorescence small animal imaging

We present a non-contact diffuse optical tomography (DOT) scanner with multi-view detection (over 360°) for localizing fluorescent markers in scattering and absorbing media, in particular small animals. It relies on time-domain detection after short pulse laser excitation. Ultrafast time-correlated single photon counting and photomultiplier tubes are used for time-domain measurements. For light collection, seven free-space optics non-contact dual wavelength detection channels comprising 14 detectors overall are placed around the subject, allowing the measurement of time point-spread functions at both excitation and fluorescence wavelengths. The scanner is endowed with a stereo camera pair for measuring the outer shape of the subject in 3D. Surface and DOT measurements are acquired simultaneously with the same laser beam. The hardware and software architecture of the scanner are discussed. Phantoms are used to validate the instrument. Results on the localization of fluorescent point-like inclusions immersed in a scattering and absorbing object are presented. The localization algorithm relies on distance ranging based on the measurement of early photons arrival times at different positions around the subject. This requires exquisite timing accuracy from the scanner. Further exploiting this capability, we show results on the effect of a scattering hetereogenity on the arrival time of early photons. These results demonstrate that our scanner provides all that is necessary for reconstructing images of small animals using full tomographic reconstruction algorithms, which will be the next step. Through its free-space optics design and the short pulse laser used, our scanner shows unprecedented timing resolution compared to other multi-view time-domain scanners.     

Analysis of digitizing errors of a laser scanning system

The digitizing errors of a high-speed 3D laser scanning system are analyzed and characterized in this paper. As the laser scanner is an electro-optical device and based on the principle of optical triangulation, the measurement accuracy is affected by the measured part geometry and its position within the scanning window. Commercial laser scanners are often calibrated in the scanning plane to account for variation of the incident angle of the laser beam. The effects of the scan depth and the projected angle, characterizing the surface normal of the measured part external to the scanning plane, on the measurement accuracy are not considered in the standard calibration process and have been identified by experiments in the present work. Experimental results indicate that the random error of the scanned data is close to the nominal value provided by the manufacturer. The systematic error shows a bilinear relationship with the scan depth and the projected angle and has a maximum value of about 160 μm. The developed empirical model correctly predicts the systematic error with a maximum deviation of only 25 μm.