基于机器视觉的齿轮多参数测量技术研究

发布时间：2018-11-20 20:06

【摘要】：齿轮作为机械传动中不可或缺的零件,历来是机械类科研人员和工程技术人员的主要研究对象之一。本文将机器视觉和图像处理技术应用于齿轮参数值和齿轮误差检测项目偏差值的测量中,详细分析了摄像机标定技术、基于图像灰度梯度的像素级与亚像素级边缘检测技术、齿轮中心定位技术以及齿轮重要参数和精度误差检测项目偏差测量技术。通过在搭建的实验平台上运用MATLAB软件实现对渐开线标准直齿圆柱齿轮多个参数值和误差检测项目偏差值的测量与获取。首先,完成背光源式齿轮参数测量平台的硬件搭建和软件功能架构。利用棋盘格标定板对摄像机的内参与外参实施了标定,矫正了拍摄图像中的畸变。其次,从伪边缘存在情况和稳定性角度对比多种像素级边缘检测算子对齿轮图像的检测效果,使用LoG算子结果图作为参与后续运算的齿轮边缘数据图。在数学形态学理论的基础上,以获得完整且封闭的齿廓图为目标,从齿轮边缘图中分割出齿轮轮齿边缘,删除图像中的杂点群并连接轮齿齿廓上的断点。再次,依据像素点灰度值分布规律,在各边缘中心点法线方向上,建立了计算各等效像素点灰度值与其到边缘中心点等效距离的算法模型。基于现有亚像素级边缘检测算法,提出在边缘点法线方向上对等效像素点灰度梯度值进行插值,纵向方向上对插值结果进行最小二乘曲线拟合,最终得到各边缘中心点亚像素级坐标值的方法。然后,借助重心法对齿轮中心进行粗定位,在分析粗定位点到齿廓上各像素点距离值变化情况的基础上,选定以齿顶圆像素点为精确定位齿轮中心的定位基准。对比三点法、二乘拟合法和本文方法的定位结果,表明本文所提出的边缘定位方法对齿轮中心的定位更为准确。最后,本文基于齿轮设计过程确定了其主要参数的测量顺序,并在标定结果的基础上获得了齿轮的多个参数值。利用本文所提的边缘检测方法和前人的插值拟合算法分别求取边缘分割后轮齿左右齿面轮廓亚像素坐标值,并在此基础上实现了对齿轮的齿距偏差、齿廓偏差以及公法线偏差的测量。根据实验齿轮的精度等级和测量对比结果,表明本文所提出的边缘检测方法及齿轮参数测量方法具有可行性和有效性。
[Abstract]:As an indispensable part of mechanical transmission, gear has always been one of the main research objects of mechanical researchers and engineers. In this paper, machine vision and image processing techniques are applied to the measurement of gear parameter value and the deviation value of gear error detection item. The camera calibration technology, pixel level and sub-pixel level edge detection technology based on image grayscale gradient are analyzed in detail. Gear center positioning technology and gear important parameters and precision error detection item deviation measurement technology. The measurement and acquisition of several parameter values and error detection item deviation values of involute standard spur gear are realized by using MATLAB software on the experimental platform. Firstly, the hardware and software structure of the backlight gear parameter measurement platform are completed. A checkerboard calibration board is used to calibrate the camera's internal and external parameters, and the distortion in the image is corrected. Secondly, from the point of view of the existence and stability of pseudo-edge, the effect of various pixel level edge detection operators on gear image detection is compared, and the result diagram of LoG operator is used as the gear edge data graph to participate in the subsequent operation. On the basis of mathematical morphology theory, with the aim of obtaining complete and closed tooth profile, the edge of gear tooth is segmented from gear edge graph, and the clutter group in the image is deleted and the breakpoints on tooth profile are connected. Thirdly, according to the distribution rule of pixel gray value, an algorithm model is established to calculate the equivalent distance between the gray value of each equivalent pixel point and the edge center point in the normal direction of each edge center point. Based on the existing sub-pixel edge detection algorithms, the gray gradient of the equivalent pixel is interpolated in the normal direction of the edge point, and the interpolation result is fitted by the least square curve in the longitudinal direction. Finally, the method of sub-pixel coordinate value of each edge center point is obtained. Then, based on the analysis of the variation of the distance between the coarse position point and each pixel point on the tooth profile, the gear center was located with the center of gravity as the precision positioning datum, and the circular pixel point at the top of the tooth was selected as the positioning datum of the gear center. Comparing the results of three point method, two multiplication fitting method and the method in this paper, it is shown that the edge location method proposed in this paper is more accurate for gear center location. Finally, the measuring sequence of the main parameters is determined based on the gear design process, and several parameter values of the gear are obtained on the basis of the calibration results. By using the edge detection method and the previous interpolation fitting algorithm, the sub-pixel coordinates of the left and right tooth profile of the rear gear are obtained, and the gear pitch deviation is realized on this basis. Measurement of tooth profile deviation and common line deviation. According to the accuracy grade of experimental gear and the results of measurement, it is shown that the method of edge detection and gear parameter measurement proposed in this paper is feasible and effective.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH132.41

【参考文献】