【作者】 黄劼；

【导师】 周肇飞；

【作者基本信息】 四川大学 ， 机械制造及其自动化， 2004， 博士

【摘要】 现代科技，尤其是航空、航天及国防等高技术领域的发展，要求制造精度更高、效率更快，因此，各种先进制造技术应运而生。制造技术与测量技术互为依托的关系决定了发展精密测量技术的必然性和迫切性，加强精密测量基础理论研究，研制满足现场需要的智能化实时测量设备是振兴我国制造业，实现从制造大国向制造强国的跨越的迫切需要。 本文以军工配套项目“球面孔系位置精密测量”为背景，开展了基于线阵CCD的机器视觉精密测量系统的研究。涉及到机器视觉光电传感器性能比较、图像采集方案实现、运动控制模式建立、图像预处理及图像识别等方面，从理论和实践上解决了用机器视觉实现曲面孔系位置测量的一系列问题，达到了每个孔位测量时间不超过20秒、光电测量部分总误差3σ＜10″的总体要求。本文的具体内容和创新有以下几点： 1．在测量方案设计上，引入“整体最优”的系统论思想，提出了通过转位控制，将三维测量转化为二维图像识别；再经过几何计算将二维平面坐标转化为球面三维孔系坐标的测量模式。从而分解了课题实现难度，在充分利用成熟技术的基础上，重点解决了图像获取和图形识别的技术难点。 2．在系统总体控制模式方面，采用了主从式多机协同操作模式：用微机作上位主处理机，完成图像识别、测量计算和人机交互；而图像采集和机械转位控制由两套单片机系统完成。该模式充分发挥了上位计算机储存量大、数学运算速度快及界面美观的优点，而四川大学博士学位论文将Windows系统难于实现的实时控制任务交由前端机处理，满足了本课题的需求，而且为类似的数据处理量大且控制实时性强的系统提供了一种可资借鉴的控制模式。 3.本文创新性地提出根据灰度直方图实现了孔型(通孔和盲孔)自动识别的构想，使系统能自动完成对不同孔型和不同孔径测量，提高了系统的智能化程度和对多孔型测量任务的适应性。 4.深入分析了机器视觉测量系统的两种常用图像传感器一一面阵CCD和线阵CCD一一所采集的图象的不同特点;剖析了扫描行距均匀性和重复性对图像识别精度的影响原因:阐述了CCD像素和光栅分辨率不同、被测圆弧各段对测量精度影响不同的机理。提出了非对称精度图像边缘点的搜索原则，实现了图像的亚像素级精度识别，从理论上解决了线阵CCD图像识别精度难以提高的难题，这对降低测量系统成本很有意义。 5.提出边缘识别与图形二值化归并处理的思想，从理论上证明了边缘点区域化搜索的可行性，以本文提出的阀值聚类法和区域搜索法构建的算法在本课题边缘识别时，比分离算法缩短时间约4秒，具有明显效果。 6.针对非对称精度扫描图像的特点，提出了曲线拟合和直线插补想结合的复合算法，可达到更高识别精度。 基于本文所述的图像采集、识别算法和三维测量转化为二维图像识别的测量模式，只需改变工件转位或移动方式，就可实现其它型面的非接触测量。本文的成果具有较广泛的适应范围，为线阵CCD在机器视觉精密测量系统中的应用提供了具有普遍意义的解决方法，对非接触精密测量技术的发展有较重要的参考价值和理论意义。关键词:球面孔系位置测量;机器视觉;线阵CCD;主从式控制模 式;图像识别更多还原

【Abstract】 Modern science and technology, especially, the development of some high-tech such as aviation, space flight, national defence and so on, require manufacture precision higher and more efficient, therefore, variety of advanced manufacturing technology appeared. The interdependent relation between manufacturing technologies and measuring technology determines that the inevitability and the urgency of the developing of precision measurement technology and strengthening the research of basic theory of precision measurement technology, developing the intelligent and real time measurement equipment to satisfy the spot, are emergency to promoting our manufacturing level and making our nation turned from large manufacturing to strong.This thesis is based on the military industry cooperative project ?Measurement of Positions for Hole Series on Curved Surface, worked on machine vision precision measure system by linear array CCD. It concerns the comparison of capacities of machine vision optical-electric sensor, implementing of image capturing, setting up movement controlling mode, image preprocessing and recognition etc. A series of problem when measuring positions of hole series on curved surface with machine vision was been solved in theory and practice,and then satisfied the total requirements that measuring time of every hole should less than twenty seconds and total error of optical-electric measurement is 3 less than 10 . The main context and innovation of this thesis are as follows:Firstly, in measurement scheme designing, brought in global optimization systematization ideology, put forward the measurement mode, which was transforming three-dimensional measurement mode to planar image recognition through locomotion control first; then, transformed planar positions to three-dimensional of hole series on curved surface by geometrical computing. Consequently, the difficulty of task was decomposed, worked at solving the technology problem of image capturing and identifying based on using mature technology adequately.Secondly, in general system controlling mode, used master-slave and multi-machine operation mode: the microcomputer was used as upper host controller to carry out image identifying, measurement data computing and human-machine communication interface; Image capturing and machine locomotion control were completed by two sets of SCM system. This mode takes the advantages of the upper host computer’s capacities such as great memory, fast computer and slinky interface into playing sufficiently, and puts the real time task which is difficult for Windows system to slave computer, upon that the requirement of the item was satisfied. Furthermore, provides a referenced control mode for some similar systems, which have a great deal of data to process and strongly real time control.Thirdly, it brought forward a creationary idea which identify hole type automatically (through hole and blind hole) according to gray-level histogram, so the system could automatically carry out measuring holes’ different type and diameter, and the intelligentizeddegree and the adaptability of measurement were enhanced.Fourthly, it analyzed different characteristic of images captured by two common image sensors -the area CCD and the linear CCD-of machine vision Measure System in depth; Anatomized the reason for the uniformity and repetition of the interval scanning effect on image identifying precision; expatiated on the effects upon measuring precision different among different CCD pixels, different raster resolution and different arc segments; put forward the searching principle of asymmetric precision image edged point and carried out sub-pixel precision recognition for image, resolved the problem theoretically that hard to advance the measuring precision of linear CCD. There is great significance to reduce the cost of measurement system.Fifthly, brought forward an ideology that merger processing the edged point recognition combining with binary image, proved the feasibility of searching the edged point by areas searching method in the t更多还原