【作者】 陈向伟；

【导师】 王龙山；

【作者基本信息】 吉林大学 ， 机械制造及其自动化， 2005， 博士

【摘要】 随着计算机视觉技术和光电技术的飞速发展,视觉检测技术得到了迅速发展和广泛应用,视觉检测技术作为一种非接触检测手段已经越来越引起人们的重视。论文结合吉林大学创新基金项目(419070200001)《基于CCD 图像的微小零件精密测量系统的研究》,从影响机械零件视觉检测应用的关键因素—检测速度、检测精度人手,在计算机视觉检测关键技术、检测系统的组成及硬件的选用、边缘检测定位的精度和视觉检测系统的标定等方面从理论上进行了深入的研究。针对机械零件边缘的特点,提出了一种基于几何特性快速圆的Hough 算法。通过实验对比可以看出,本文提出的改进的圆的Hough 算法在运行效率上比标准圆的Hough 算法提高了一个数量级;为提高直线边缘的定位精度,提出了用于直线边缘的亚像素定位方法——最小二乘回归直线拟合法。该算法速度较快,具有较高的定位精度,适应于直线边缘的整体定位,为工业应用打下了理论基础;深入分析了镜头产生畸变的原因,提出了利用平面网格对视觉检测系统的几何畸变误差进行校正的基本原理。实验结果表明,采用本文提出的校正方法对图像径向和切向畸变校正效果有明显改善。并根据二维图像测量实际应用的特点,提出了一种简单、便捷的标定方法。该方法可同时消除镜头畸变误差和标定过程引入的随机误差。将研究结果应用到复杂结构尺寸检测——石墨电极锥接头螺纹几何参数的测量和微小结构尺寸检测——传动用滚子链链板节距和微小零件圆度误差的测量中,构建了视觉检测系统的硬件结构,开发了螺纹几何参数测量软件,对石墨电极锥接头螺纹的牙型半角、锥度和节距等几何参数进行了定量测量。推导了螺旋升角对牙型半角变化量的影响公式,为误差修正奠定了理论基础。通过测量实验可以看出:本系统的测量结果是令人满意的。论文所作的研究与开发工作对提高计算机视觉检测技术水平具有一定的理论意义和工程应用价值。更多还原

【Abstract】 With the development of computer technology and photoelectricity, a new inspection technology based on computer visual has been presented. The visual inspection is an inspection method which treats image as information means or carrier of inspection and transfer when objects are inspected, and the goal of this method mean that picking useful signal up from image. It is a integrated modern inspection technology including photoelectricity, computer graphics, information management and computer visual etc. and based on modern optics. The apparatus based on visual inspection technique can realize intelligence, digital, miniaturization, network and multifunction and has the capabilities of inspection on real-time, real analysis and control, which is used and attended widely in the fields of war industry, medicine, etc. Computer vision has wide sensitive scale for action spectrum,high measure precision and steady inspection result, so it contributes to many fields to enhance production efficiency, ensure product quality and enhance product precision etc. greatly. From 1980s the international application of visual system has not been interrupted. Computer vision has been developed brilliantly and widely used in every aspect with the continual appearance of new concepts, theories and methods. In our country, industry visual system is being in the process of concept-introduction, and after every field’s advanced enterprises have solved production automation, they moved the view to automatic inspection. But compared to overseas, there are at least 20 years’gap due to the level of domestic computer visual inspection R&D. So it is very important to more develop theory & practice research of visual inspection, to absorb advanced foreign technique & science thinking, and to extend the fields using visual inspection. This Ph.D. dissertation mostly studies the computer visual inspection method of mechanical part’s geometric dimension and shape error. Combining the project of Innovation Foundation for Jilin university(419070200001): Research of precision measurement system of weenie part on CCD image, there doesn’t often need quite complicated arithmetic but high precision because of more prior knowledge & measured surface measured and optics characteristic varied unconspicuously or regularly in the process of producing part. Therefore this paper goes deep into the research of computer visual inspection technology mostly from some key factors that affect visual measure, such as measure speed and measure precision. And the result is used in complicated structure’s dimension inspection—geometry parameter measurement of truncated cone nipples of electrode’s taper screw thread and weenie structure dimension inspection—the measure system of transmission roller chain’s link joint pitch and round error of weenie parts. So the study is a typical model combining computer visual inspection technique to production practice and has highly production application value and wide applicability. There are introduction of paper studying according to chapter in below. Chapter one—Introduction. In this chapter, in the base of plentiful abroad and domestic information on computer visual inspection technique, the paper discusses simply development process of computer vision, summarizes abroad and domestic research actuality and trends of computer vision inspection technique and introduces the appliance results of computer vision inspection technique. In the paper, the image measure’s value of theory and practice based on computer vision is emphasized. And key technique is clarified. At last the author makes primary study contents of this paper clear. Chapter two—The inspection system of mechanical part. In this chapter, the character and advantage of computer visual inspection are analysed, and the structure of computer visual inspection system is discussed, in the end general principles of choosing hardware of computer visual inspection system are given and the necessity of studying computer visual inspection is emphasized. Chapter three—The analyzing and processing of the image. In this chapter, the theory analysis of image processing in visual inspection is discussed, which discusses general image preprocess methods using computer visual inspection technique to inspect and identify inspected objects from two aspects of edge detection and image segmentation. In the aspect of image segmentation, P parameter method,maximum variance threshold method, double-humped histogram method, partition threshold method, minimum error split method and multithreshold method are mostly discussed. In the aspect of edge detection, first-order differential edge detection operator, second-order differential edge detection operator, Canny edge detection operator, method of edge detection based on wavelets transform and line & circle’s Hough transformation are studied and discussed, the merits and demerits of first-order differential edge detection operator and second order differential edge detection operator are analysed. A fast circle Hough arithmetic based on geometrical property is proposed. Through the experiments comparison we can see that the run efficiency of reformative circle’s Hough arithmetic proposed by this paper is advanced one order than standard circle’s. Chapter four—Sub-pixel edge inspection. The technique of sub-pixel edge inspection is a efficient method which use software arithmetic to advance measure precision, so it stays important position in image precision measurement. In chapter four, the sub-pixel allocation theory based on computer vision in image measure is dissertated. In the base of discussing sub-pixel edge inspection arithmetic of multinomial interpolation, cubic spline function interpolation, gray moments edge orientation method and space moments edge orientation method, a sub-pixel orientation method used to line edge—least-squares fitted regression line is proposed, Contrast test is done to some sub-pixel edge detection arithmetic’s orientation precision and running time, which has faster speed of running and better orientation precision and is adaptive to line edge’s whole orientation and makes a theory base for industry application. Chapter five—The calibration of computer visual inspection system. Calibration technique is a key factor affecting the precision of visual inspection system. In chapter five, calibration technique in computer visual inspection is summarized simply and basic theory using plane grid to proofread the geometry aberrance error of visual inspection system. The experimental result shows there has obvious improvement to image radial and tangent aberrance emendation using the method proposed by this paper. According to the practical application character of 2-D image measure, the paper proposes a simple & enchiridion calibration method. It is according to visual field of different orientation, doing many times calibration and更多还原