【作者】 李哲林；

【导师】 夏琴香；

【作者基本信息】 华南理工大学 ， 机械制造及其自动化， 2013， 博士

【摘要】 长轴类大锻件是制造重大装备大型关键零部件的基础，轴向尺寸较大的锻件都采用自由锻的方法锻造，但自由锻加工时锻件尺寸精度难以控制，需要保留较大余量。目前国内大部分锻造企业采用人工测量方法来获得锻件尺寸，存在危险大、测量参数少等问题。基于CCD视觉技术的测量方法可实现快速、非接触、多参数和高精度测量，是当前研究的热点和趋势。高温锻件具有辐射度大且变化大等特点，导致CCD采集高温锻件图像、获取图像上特征、图像匹配及三维结构重构等关键技术需要解决。本文对CCD图像测量高温长轴类锻件的关键技术进行了研究，其主要研究内容和成果如下：（1）针对高温锻件体发射大量的热辐射和成形过程中温度变化大，分析高温锻件辐射出射度与波长和温度的关系，研究了一种基于基色灰度的自适应曝光模型，获得灰度值稳定的单基色灰度图像；然后分析了图像中噪声的特点，并对中值滤波和高斯滤波算法的滤波特性进行研究，对图像降噪算法进行了研究，提出先用中值滤波消除图像中的椒盐噪声然后用高斯滤波降低随机噪声的组合滤波算法，使得图像信噪比PSNR值明显提升。（2）为从图像中分割出锻件目标，分析高温图像中三种基色灰度的规律，分别对图像区域特征和边缘特征进行分析，发现在不同温度区域中基色灰度具有明显比例相关性特点，提出采用红色与蓝色灰度比做为动态相对阈值识别目标和背景；研究高温锻件图像中锻件边缘的灰度变化规律，采用ANSYS对高温锻件边缘温度场进行数值模拟，提出用高温区灰度平均值作为边缘分割阈值；研究了基于亚像素插值方法的边缘提取算法，可有效提高远距离采集的锻件图像的分割精度。（3）分析双目视觉测量的基本理论和方法，设计一种针对轴类锻件测量的摄像机上下布置模式测量系统，并建立竖直外极线双目测量系统的空间模型，设计大场景用标定板，标定出了两个摄像机的内参数矩阵；通过计算两个CCD的空间位置，得出上下摄像头空间平移和旋转关系，建立外极线约束方程，并采用极线校正方法对采集的双目图像进行校正，通过试验验证了这种系统的可行性。（4）为解决高温锻件图像上特征匹配的问题，分析了轴类锻件成型过程中结构形状的变化特点及图像内部特征；研究了基于锻件截面线的特征提取方法，通过提取截面线的灰度分布曲线上局部极大值作为特征点，建立了有效的稀疏特征矩阵；提出结合外极线约束，通过欧式距离对两幅图像中的对应扫描线上稀疏特征点进行匹配，可得到特征点匹配矩阵，并求解出两幅图像的匹配点的视差矩阵。（5）高温锻件三维重构和测量是本研究的目标。研究从截面线上空间离散点拟合圆及锻件中心线拟合方法，并研究了锻件三维模型重构和坐标系转换算法；在锻造车间的试验结果表明，该高温锻件测量系统无结构光或激光测距仪等辅助器件，同样可以从高温锻件图像重构出锻件三维图形，且测量精度和速度指标能满足生产实际需求；还设计研制了一套双目测量系统和开发了一套锻件尺寸测量软件，可实时监控测量结果，对比锻件的测量重构模型和理论模型的差异。更多还原

【Abstract】 Long-shaft Heavy Forging is the foundation of large components manufacturing ofimportant equipment. Large axial dimensions shaft are forged by free forging method. But itis difficult to control precision dimensions of forgings. The large machining allowance is keep.At present, manual measurement method is used in most of forging firms in order obtain thesize of forging. Measurement Method based on CCD image is a fast, non-contact, multi-parameters and high accuracy measurement. It is the current hot topics and trends of the study.The large radiation of high temperature forgings causes many key technologies, such as imageacquisition, features, image matching,3D reconstruction. Some key technology will beresearched on CCD image measurement of long shaft forgings in this article. The maincontents and results of the study are as follows:(1) For large number of hot radiation and large range of the temperature in formingprocess of high temperature forging, the relationship among radiant exitance and wavelengthand temperature are analysed., an automatic exposure model is studied based on the basecolor grey of image. The stability single grey value of grey image can be obtained by thismethod. Features of noise in images are analysed. Characteristic of the medium value filterand Gaussian filter are researched. First pepper noise is eliminated with medium value filterand random noise is reducing with Gaussian filter by combination filter algorithm. The PSNRvalue of images decreased obviously.(2) In order to obtain forging target from images, law of three basic color grey in imageare analysed, image regional features and edge features is respectively analyzed. Thecorrelation between basic color grey in different temperature zone is found. The dynamicrelative threshold from the red and blue grey ration is used to recognize target and background.The grey change law in the forging edge of temperature forging image is studied. Edgetemperature field of high temperature forging is simulated by using ANSYS software. It isfound that the average grey value of high temperature area can be as edge segmentationthreshold. A sub-pixel interpolation algorithm is studied. It can be effective improvesegmentation precision for far distance collection forging image.(3) The theory and method of stereo vision measurement is analyzed. A measurementsystem layout mode of up and down camera is designed for measurement of axis class forging.The vertical epipolar line stereo vision measurement system is modeled. A big calibrationboard is made for big scene measurement. Intrinsic parameters matrix of two CCD cameras are obtained by calibration. By calculating space location of two cameras, the spacetranslation and rotating in external parameters matrix of up and down camera is obtainedrelationship. The epipolar line constraints equation is formulated. The stereo images arerectified by using equation. The experiment results show that this model is feasible.(4) In order to realize features match problem of high temperature forging image, theshape change characteristics and internal features of the image are analyzed. A featureextraction method based on section line is studied. The extremum points in grey distributioncurve are be as features points. A sparse features matrix is established. By Euclidean distanceand epipolar line constraints, sparse feature points in the section line from stereo images arematched. A match matrix and disparity matrix of features points are obtained by this method.(5)3D reconstruction and measurement of high temperature forging are the goal of thisarticle. Fitting method of space discrete point on section line and the forging center line arestudied.3D model reconstruction of forging and coordinate system conversion algorithm arestudied. The experiment results from forging workshop show that3D model could bereconstructed from high temperature forging image without structure light or laser ranginginstrument such as auxiliary devices. The measurement precision and speed can meet thedemand of actual production. Measurement system device and measurement software aredesigned and developed. It realize real-time monitor the dimension of forging and analyzedifference between the reconstruction model and theoretical model of forgings.更多还原