基于机器视觉的数控自动编程系统开发

基于机器视觉的数控加工技术综合运用机器视觉技术和数控加工技术,直接利用视觉图像数据实现数控加工编程。将机器视觉技术应用于数控加工编程技术中,构造出基于机器视觉的数控自动编程系统。通过机器视觉采集系统来获取工件图像,并对图像进行处理得到工件图元的边缘轮廓。进而将加工的边缘轮廓进行矢量化处理得到数控加工轨迹,并最终根据加工工艺参数的设定转化成数控加工程序,实现了工件的自动编程。     

工件轮廓曲线磨削加工精度的原位测量与评定方法研究

为了解决曲线磨削中工件轮廓加工精度的在位评定问题,提出了一种基于机器视觉的工件轮廓曲线磨削加工精度原位测量和评定方法。应用这一方法,通过视觉测量系统和亚像素边缘提取算法,得到工件的实际轮廓边缘点。通过轮廓匹配算法来确认工件实际轮廓与理论轮廓的最优匹配位置,进而计算轮廓偏差,进行加工精度评定。分别进行了标定板试验和工件试验,表明所提出的评定方法精度可达0.6像素,能有效保证工件轮廓在位检测的精度和效率。     

齿轮倒角轮廓协同高精度测量方法的研究

为实现对制造业中复杂工件(如汽车变速箱中的倒角齿轮)表面轮廓的快速、精密测量,研制了协同视觉表面轮廓测量系统。对该系统中采用的图像处理、测量路径的规划进行了研究。首先使用面阵图像传感器拍摄齿轮图像,对图像进行边缘处理,利用齿轮的边缘点拟合齿轮的中心,在结合标定算法定位齿轮倒角的待测局部区域。然后,根据测量要求规划测量路径,驱动XY电动平台携带高精密的点视觉激光位移传感器沿测量路径完成对带测量区域的测量。最后,使用计算机对激光位移传感器所测得数据进行处理,输出测量报告。该系统经工厂实际测量验证,对齿轮倒角表面轮廓的高度方向(Z方向)测量精度为1μm,对X和Y方向的测量精度达到5μm以上。结合图像传感器测速快,激光传感器精度高的优点,提出一种利用图像传感器实现高速定位,激光传感器实现精密测量的协同高精度视觉测量方法,并将其用于了齿轮倒角轮廓的测量。     

基于双目视觉的工件尺寸在机三维测量

由于传统测量方法及仪器设备无法满足在机实时检测工件三维尺寸的要求,本文基于双目立体视觉视差原理,搭建了一套可在车间现场操作的在机测量工件三维尺寸的视觉系统.首先,介绍了改进的基于平面圆靶标的双目视觉测量系统标定方法.然后,用阈值分割算法从背景中识别出待测工件,采用Canny算子和多边形逼近法提取工件轮廓和轮廓关键特征点,并在极线约束的特征点匹配算法基础上,提出一种基于灰度相关的密集型精匹配算法,获得了工件边缘轮廓关键特征点云的三维坐标数据.最后,对匹配所得特征点云进行数据处理和特征拟合计算,得到了工件的关键尺寸.在车间数控机床上对有斜面及孔结构的工件进行了在机测量实验,结果显示,该系统检测精度达±1.3％以上.     

基于定位平台的大尺寸工件视觉测量技术研究

对于大幅面工件的几何尺寸很难进行全尺寸视觉测量的问题,提出一种单目视觉加运动控制联合的测量方法.该方法结合了图像检测功能与运动平台的可移动特点,对大尺度工件进行分区编号,控制直线精密定位平台按逆时针顺序对相应区域进行图像采集处理,并建立一种基于Canny边缘检测与分段序贯最小二乘法拟合求交点的角点检测算法,通过获取先后两不同区域图像角点坐标的偏差,结合定位平台位移信息,实现对大幅面工件的全尺寸视觉测量.最后运用建立的大尺寸工件视觉测量系统对特征丰富的工件进行测量实验,实验结果表明系统在200mm测量范围内可保证0.05mm的精度要求.     

基于机器视觉的工件自动分拣系统的研究

介绍了机器视觉技术在工件识别中的应用以及主要图像处理算法。采用机器示教方法建立待处理零件的特征库,通过图像预处理、阈值分割、边缘和轮廓提取算法等获取零件图像的特征并与标准模板比较,从而判别待测零件的合格性。为解决零件摆放的位姿偏差,采用边缘检测和轮廓跟踪算法确定工件形心位置和旋转角度,根据与模板零件的相似性测度判断工件的特性,为自动分拣提供决策支持。实验验证了本算法的有效性。     

基于机器视觉的大型换热器管材孔系测量

针对大型换热器管材孔系，提出基于精密运动控制的机器视觉测量方案。运动控制系统携带相机完成工件全局扫描。机器视觉系统采集工件局部位置图像，使用高斯滤波法、最大类间方差法和Canny边缘检测方法自适应地检测孔系边缘，采用Hough梯度法提取单孔轮廓；利用优化的最小二乘拟合法测量单孔的孔心坐标和孔径，计算相邻孔的孔桥距。对测量方法进行实验验证，实验结果表明，孔径偏差低于0.031 mm，孔桥距偏差低于0.032 mm，满足测量精度要求。     

三维机器视觉测量系统图像边缘提取算法研究

在三维机器视觉测量系统中,二维数字图像边缘提取是进行三维坐标计算的必要前提。对几种常用边缘检测算子的边缘提取效果进行了比较;提出了一种基于Priwitt算子的改进边缘提取算法;给出了三维机器视觉测量系统中激光线图像与零件特征边缘图像交点的求解方法,为三维机器视觉测量系统的三维坐标计算奠定了基础。     

基于机器视觉的螺纹牙型角检测方法研究

针对螺纹牙型角测量过程中自动化程度和检测效率低等问题,设计了一种基于机器视觉的螺纹牙型角测量系统,开发了螺纹检测平台控制算法和图像处理算法。该系统采用工业相机获取螺纹图像,利用边缘增强的Otsu算法分割螺纹工件图像,基于分割图像的LSD算法检测牙型角,测得牙型角的侧边缘。分析螺纹图像边界与实际轮廓的关系,基于螺纹牙型与图像阴影区域边界差值方程校正测量结果。采用标定算法标定相机,得到相机径向畸变系数和内参数矩阵。螺纹牙型角检测实验结果表明,系统测得的牙型角与实际工具测量牙型角均值相差0.022°,可知,本系统能较准确地测量出螺纹牙型角,因而可以应用于螺纹牙型角的准确测量。     

基于机器视觉的共轨针阀尺寸参数在线测量

针对喷油嘴共轨针阀零件的尾部尺寸参数在线检测需求,开发了一套基于机器视觉技术的共轨针阀多参数尺寸在线测量系统。通过对共轨针阀零件尾部图像采集,引入基于中值滤波的小区域滤波算法,大大缩短了滤波降噪时间;在轮廓边缘检测方面,通过对比实验,引入了基于Sobel的小区域定向搜索边缘检测算法,提高了边缘定位效率,并采用小区域边缘特征点跟踪方法,基于学习训练逻辑,完成目标特征点定位;在边缘亚像素定位和拟合方面,引入基于稳健线性回归的直线拟合方法,在特定小区域下完成高精度边缘拟合,并完成直径、角度、距离的多参数同时计算,系统测试表明,该视觉系统精确稳定,具有良好的应用。     

全自动太阳跟踪器研制和应用

本文着重描述了全自动太阳跟踪器的原理、结构、主要功能和应用。全自动太阳跟踪器具有两种可自行切换的跟踪方式：传感器跟踪方式和太阳运行轨迹跟踪方式。这两种方式自行切换,互相配合,实现了高精度的全天候太阳的自动跟踪。全自动太阳跟踪器的研制成功,使在没有改进传感器的情况下,可提高直接辐射、总辐射的测量准确度,解决了直接辐射表信号线绕线的难题,并实现了直接辐射的全自动测量。由于全自动和准确跟踪太阳,可进行日照时数的自动测量,故填补了我国在日照时数自动测量方面的空白。另外,还可进行散射辐射的自动测量和一些其它应用。     

用动态水平基准测量系统实时测量船体横扭角和纵挠角

设计了动态实时水平基准测量系统,讨论了用该测量系统测量远望号测量船变形角的可行性及利用该系统测量船体横扭角和纵挠角的原理和误差项。依据测量船的实际需要确定了动态水平基准测量系统的关键技术指标,并分析了测量船体横扭角和纵挠角的测量精度。分析结果表明：系统对船体横扭角和纵挠角的测量误差均不大于3″,测量距离理论上不受任何限制,远大于200 m(测量船船体长度)。利用现有外场试验条件进行了码头停靠状态下远望号测量船两测量点间的船体横扭角和纵挠角测量试验,结果表明：利用动态水平基准测量系统实时测量船体横扭角和纵挠角的方法切实可行,不仅提高了测量精度,而且克服了传统的大钢管基准测量法体积大、重量沉,装调及维护困难等缺陷。实验显示,该动态实时水平基准测量系统的应用领域相对广泛。     

基于DSP的变频电源电参数测量系统

变频电源越来越多地用作标准供电电源，为电器提供优良的供电环境，便于客观考核用电器的技术性能。设计了变频电源的电参数测量系统，以TMS320F2812芯片为核心并结合高精度16位芯片ADS8364来实现；在时序方面采用CPLD来控制，同时实现了频率的跟踪和同步采样功能。最后结合实际电路，进行了系统软件方面的设计和仿真。实践表明：该设计完全可以达到预期的采样精度，满足对变频电源电参数测量的需要。     

‘Open-loop’ tracking interferometer for machine tool volumetric error measurement—Two-dimensional case

The tracking interferometer, or the laser tracker, is a laser interferometer with a steering mechanism to regulate the laser beam direction to follow a retroreflector (“target”). Applying the multilateration principle, it measures the target's three-dimensional position at an arbitrary location in the workspace. Its application to the volumetric accuracy measurement for coordinate measurement machines or machine tools has been long studied. In this paper, we propose the ‘open-loop’ tracking interferometer, where the laser beam is regulated toward the command target position. This eliminates the automated tracking mechanism and thus may significantly reduce the manufacturing cost of conventional tracking interferometers. The objective of this paper is to validate this ‘open-loop’ tracking interferometer concept by investigating its measurement uncertainty both experimentally and analytically. To simplify the problem, this paper focuses on the measurement of the target's two-dimensional position by using a single-axis ‘open-loop’ tracking interferometer prototype.     

动态视角下自主目标识别与跟踪

针对动态视角下由于相机高频率晃动导致的常用目标识别及跟踪算法准确率较低的问题,提出了一种基于Canny和GrabCut的自适应窗口式目标跟踪算法。首先使用加速稳健特征（SURF）算法学习图片库并记忆图片特征,设计基于SURF算法的记忆库目标识别算法;然后,对上述目标区域采用GrabCut的自适应优化算法进行感兴趣区域分割,实现目标粗略跟踪;最后,设计基于Canny算法的窗口式算法进行目标精确追踪。实验结果表明,所设计的算法能快速地识别目标、精确地勾勒出其轮廓并且稳定跟踪目标,相比其他算法,算法在实时性和精确性方面有显著提高。     

光电跟踪系统鲁棒自调整内模控制设计与分析

光电跟踪系统广泛应用于制导武器、航空航天观测设备以及靶场光电测量仪器等领域。针对光电成像测量时滞和模型不确定性,设计了一种鲁棒自调整内模控制算法。给出基于相位裕度和幅值裕度指标的系统控制器整定方程,分析得到了模型失配情况下的系统鲁棒稳定条件以及参数的设计准则,引入基于二次型性能指标的优化方法在线自适应调整控制滤波器参数来提高系统的控制性能和鲁棒性。在光电跟踪转台上的实验结果表明,所设计算法可有效地克服光电图像跟踪器测量时滞所造成的跟踪偏差,跟踪精度小于1密位,保证了成像传感器对目标的快速准确跟踪。     

Measuring large 3D structures using four portable tracking laser interferometers

A new concept that allows measuring 1D–3D objects in the range of several centimeters to 5 m × 5 m × 5 m is presented. In general terms the concept can be seen as a task specific correction of geometrical errors of coordinate measuring machines (CMMs). The developed system comprises a commercial CMM, its measurement and its evaluation software and a set of at least four high accurate tracking laser interferometers. The CMM is simply used as a mover which allows to capture points on the surface of a measuring object. In parallel the tracking laser interferometers follow a retro-reflector located close to the stylus tip of the tactile probe of the CMM. Based on a multi-lateration algorithm 3D-positions are calculated from the measured interferometric distances. Finally, two sets of coordinates emerged, namely, one by the CMM and the second from the metrological frame of the tracking laser interferometers. The interferometrically measured positions are usually more precise than the positions measured by the CMM. This is due to the high accuracy of the interferometric system and also due to the fact that the measurement positions are taken in a manner which almost avoids Abbe errors. Because of that, the measurement positions of the CMM are substituted with the more accurate measurement points calculated from distance measurements of the tracking interferometers. The position coordinates thus obtained are used for the further computerized evaluations, which yield the geometric parameters of the object measured. First measurements under laboratory condition show very promising results. It has been demonstrated that the concept is suitable for the high precision calibration of large workpieces with small tolerances, for instance, for the calibration of large gears for the windmill industry.     

Exploiting robotic link deflection for low-cost force measurement in manufacturing

A custom, low-cost force measurement system for manufacturing is presented that exploits robotic link deflection for measurement purposes instead of utilizing a commercially purchased load cell. Force measurement can be an important aspect to many manufacturing processes, as it can allow for force feedback control or other process-planning related operations, such as tool-workpiece autozero or through-the-tool joint tracking. This system is demonstrated on a machine that is used for Friction Stir Welding research, and a Finite Element Analysis is also conducted on the robotic link that is used for axial force measurement purposes. This system may be particularly attractive to small businesses or low-volume manufacturers whose smaller operating budgets may normally prohibit them from implementing force measurement systems.     

基于八邻域跟踪算法的封闭区域几何尺寸测量

提出了基于八邻域跟踪算法的任意形状封闭区域几何尺寸测量。首先对目标图像进行二值化．然后再利用八邻域跟踪目标区域边缘得到目标边缘坐标点,最后根据欧氏距离公式和格林公式计算目标区域的周长和面积。该方法实现容易、计算结果准确、测量精度较高,为图像的进一步分析提供了依据。     

A technique for the measurement of patellar tracking during weight-bearing activities using ultrasound

A new fixation device, the femoral clamp, was developed in this study for the ultrasound measurement of patellar medio-lateral motion during sitting and squatting knee flexion/extension. Seventeen subjects, 6 males, 11 females, aged between 18 and 40 years were recruited for the test. Results showed that the patella moved medially then laterally from extension to flexion when sitting. Weight-bearing knee motion produced a more laterally tracked patella without the presence of the initial medial patellar translation. The tracking patterns of the patellae were similar regardless of knee movement direction. The patellar lateral position was greatly affected by the movement task (p < 0.0005), and was also influenced by gender for maximum medial position (p < 0.05). The reproducibility of the measurement was between 0.29 and 0.90 for the intra-rater and 0.34-0.75 for the inter-rater reliability. The accuracy of the ultrasound measurement was validated by interventional magnetic resonance (iMR) imaging of the patella and the mean error of the measurement was 1.4 +/- 3.2 mm. Although further research is needed to improve the accuracy and reliability of this method, it has demonstrated the feasibility of obtaining patellar tracking data during load-bearing activities.