基于机器视觉的孔组直径测量技术研究

发布时间：2018-01-13 18:39

本文关键词：基于机器视觉的孔组直径测量技术研究　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：基于机器视觉的零部件非接触式尺寸检测以其检测的速度快,精度高,适用性强的特点,被广泛应用于工业自动化生产中。在制造业中,圆孔是机械加工零件上普遍存在的重要特征,其尺寸加工精度直接影响着孔轴的配合质量,进而对系统的运转性能和安全性能产生重要影响。本文以机器视觉为基础,提出了一套以法兰盘为代表的孔组类零件内径尺寸的测量方案。论文的研究内容主要涉及以下几个方面:边缘检测是图像处理和机器视觉测量中最基本的步骤之一。在本文中,通过Canny算子对边缘进行粗定位,得到圆孔的像素级边缘。再结合以像素点为中心的3?3窗口内其他像素点的灰度信息,计算出边缘点的亚像素坐标。本文采用改进的多项式插值算法来进行圆孔边缘点的亚像素定位,与传统多项式插值算法相比,改进的算法增加了两个方向模板,在边缘点的灰度梯度方向上进行多项式插值计算,提高了圆孔边缘的亚像素定位精度。本文对Hough变换法、形心法、灰度重心法以及最小二乘拟合的圆心检测算法进行了分析,提出运用计算效率更高的弦端点法来进行圆心坐标的检测和圆孔直径的计算。对弦端点法的原理进行了分析,并说明了弦端点法中大误差数据点的来源及剔除方法。考虑到相机成像面和待测平面存在夹角,使圆孔图像的采集因透视投影畸变产生误差,本文将检测到的圆孔边缘点的亚像素坐标投影到世界坐标系中,在法兰盘平面进行圆心坐标和圆孔直径的计算。通过以上分析,建立基于机器视觉的孔组直径测量模型,将世界坐标系建立在标定板平面上,根据在标定面平面标定得出的参数,建立世界坐标系和摄像机坐标系间的位置关系,进而推导出待测面的平面方程。将求出的边缘点的摄像机坐标,转换到世界坐标系中,令圆孔直径的计算问题简化为在待测平面进行的二维问题。基于对以上内容的研究,搭建孔组直径的测量实验平台。对提出的测量方案进行实验验证。将本文测量方案的实验结果与法兰盘各孔直径的千分尺测量值进行对比。在正反标靶的对比实验中,反面标靶的测量精度更高,其测量的平均误差低于正面标靶测量的平均误差0.095mm。弦端点法和拟合法的对比实验结果显示,弦端点法的检测精度更高,孔径检测的平均误差为0.055mm。本文在实验室环境下,进行了基于机器视觉的孔组直径测量研究,对于非接触式测量和机器视觉技术在工程中的应用具有一定的探究意义。
[Abstract]:Non-contact dimension detection of parts based on machine vision is widely used in industrial automation because of its high speed, high accuracy and strong applicability. Circular hole is an important feature of machining parts, and its dimension machining accuracy directly affects the quality of the matching of the hole shaft. This paper is based on machine vision. This paper puts forward a set of measuring scheme of inner diameter of hole-group parts represented by flange. The research contents of this paper mainly involve the following aspects:. Edge detection is one of the most basic steps in image processing and machine vision measurement. The pixel edge of the circular hole is obtained by rough location of the edge through the Canny operator, and the pixel center is combined with 3? 3The gray level information of other pixel points in the window, and the sub-pixel coordinates of the edge points are calculated. In this paper, the improved polynomial interpolation algorithm is used to locate the sub-pixel of the circular hole edge points, which is compared with the traditional polynomial interpolation algorithm. The improved algorithm adds two direction templates, carries on the polynomial interpolation computation in the edge point grayscale gradient direction, improves the circle hole edge sub-pixel localization accuracy. This paper deals with the Hough transform method, the centroid method. The gray center of gravity method and the least square fitting centroid detection algorithm are analyzed. The chord endpoint method with higher computational efficiency is proposed to detect the center coordinates and calculate the diameter of the circular hole. The principle of the chord end point method is analyzed. The source of the large error data points in the chord endpoint method and the method of eliminating the large error data points are explained. Considering the angle between the camera imaging plane and the plane to be measured, the acquisition of the circular hole image is caused by the perspective projection distortion. In this paper, the subpixel coordinates of the detected edge point of the circular hole are projected into the world coordinate system, and the center coordinates and the diameter of the circular hole are calculated in the flange plane. The measurement model of hole group diameter based on machine vision is established. The world coordinate system is built on the plane of calibration board and the parameters are calibrated in the plane of calibration plane. The position relation between the world coordinate system and the camera coordinate system is established, and the plane equation of the measured plane is deduced. The camera coordinate of the edge point is converted to the world coordinate system. The calculation of the diameter of a circular hole is simplified to a two-dimensional problem in the plane to be measured. Set up an experimental platform for measuring hole group diameters. Test verification of the proposed measurement scheme. Compare the experimental results of the present measurement scheme with the micrometer measurements of the diameter of the flange holes. Test. The measurement accuracy of the reverse target is higher than that of the front target. The experimental results show that the average error of the reverse target is lower than that of the front target. The comparison between the chord end method and the fitting method shows that the accuracy of the method is lower than that of the front target. The accuracy of the chord endpoint method is higher, the average error of aperture detection is 0.055 mm. in this paper, the diameter measurement of hole group based on machine vision is studied in laboratory environment. It has certain exploration significance for the application of non-contact measurement and machine vision technology in engineering.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG81;TP391.41

【相似文献】