【作者】 胡而已；

【导师】 陈传尧； 何玉明；

【作者基本信息】 华中科技大学 ， 固体力学， 2009， 博士

【摘要】 投影条纹相位测量技术是实验力学和光学无损检测中常用的一种非接触式三维形貌及表面缺陷测量方法，目前主要用于测量静止物体的三维形貌。为使投影条纹相位测量技术应用于运动物体的三维形貌及表面缺陷测量中，分别构建了基于面阵和线阵CCD相机的测量系统，采用相移法和傅里叶变换法提取变形条纹图的相位，成功重建了物体三维形貌。应用线扫描图像系统对物体的动态变形也进行了实验研究。具体研究内容如下：论文提出了一种改进的基于相移技术的运动物体三维形貌在线测量方法。采用普通面阵CCD相机搭建了测量系统。循环投影相移数字条纹图到待测运动物体表面，通过控制投影与图像采集时间间隔，可以拍摄多幅具有相移量相同的物体表面条纹图，为获得物体的三维形貌对传统的相移算法进行了改进。该方法成功地将相移技术用于运动物体三维形貌测量，保证了表面曲率较大处的物三维体形貌的测量精度，且系统硬件要求较低，适合在线产品质量检测。论文详细分析了相移中引入相位误差的主要因素，包括图像量化误差、投影与采集系统非线性误差和图像饱和误差，分别研究了不同因素产生的相位误差特征，针对不同的误差因素研究了相应的相位误差抑制方法。(1)理论推导了图像量化导致的相位误差的标准差公式，采用离散的正弦信号模拟采集获得的条纹图，理论分析结合数值模拟得出，图像量化导致的相位误差随着相移步数、量化位数、有效量化系数和条纹调制度的增加而降低。(2)投影和图像采集系统的非线性响应会在相位图中引入周期性的波动误差。分析总结了非线性响应引起相位误差的理论公式，得出三步相移算法中相位波动误差的空间频率为投影条纹频率的四倍，而四步及四步以上相移算法可有效地抑制系统非线性误差，且实验结果与理论分析吻合。(3)条纹图像的局部饱和同样可在相位图中产生周期性的相位波动误差。首先定义了饱和系数K来描述图像的饱和程度，在提取不同饱和程度的正弦模拟信号模拟条纹的相位后发现，饱和导致的相位误差的标准差随着饱和系数的增大而增大。针对图像饱和导致的相位波动误差，改进了饱和误差抑制算法，给出了基于三步、四步和六步相移技术的饱和误差抑制算法的公式。模拟和实验结果表明，饱和系数在算法适用范围内时，基于N帧相移技术的饱和误差抑制算法可以有效地消除该类的相位误差；饱和误差抑制算法的相位修正范围将随帧数N的增加而拓广。论文研究了基于线阵相机扫描技术的变速运动物体三维形貌测量方法，搭建了图像测量系统。投影数字正弦条纹图对待测运动物体表面进行光学编码，采用线扫描相机逐行扫描获取变形条纹图。通过旋转速度编码器实时获得物体运动速度，反馈给计算机控制线扫描相机的扫描频率，解决了变速运动物体成像的畸变问题。分别采用提取条纹中心线法和傅里叶变换法提取物体三维形貌，讨论了两种方法的优缺点。比较分析不同形貌提取方法的实验结果可得，提取条纹中心线法的测量精度一般要低于傅里叶变换法，但傅里叶变换法测量表面曲率较大的物体时误差较大，这两种方法在实际应用中可以互为补充。所研究的双频条纹傅里叶变换法则在保证物体三维形貌测量精度的同时可有效地克服表面阶跃引起的相位包裹。论文最后研究了基于线扫描图像系统的弦线振动频率和物体动态变形测量。采用线扫描系统对振动的弦线进行动态成像，获得弦线上某一测点的位置变化曲线，由傅里叶变换法提取基频信息，并与理论值进行比较来验证该方法的可靠性。物体动态变形测量系统与运动三维形貌测量系统类似，不同点在于测量动态变形时待测物体与图像系统间无相对运动，线扫描相机反复扫描待测物体表面上同一行，获得该行在不同时刻的变形条纹信息，采用傅里叶变换法提取该行的轮廓线序列，从而比较被测物体的动态变形量。本文采用该方法测量了人体小臂在肌肉张弛过程中的动态变形。更多还原

【Abstract】 Projection grating profilometry has been widely used in experimental mechanics, 3-D sensing, machine vision, robot simulation, industry monitoring, biomedicine, etc. because of the advantages of full field measurement, high speed measurement and high resolution. In this work, we use a frame CCD camera and a line-scan CCD camera to construct image detection systems to detect the 3-D profile and surface defect of a moving object, respectively. In addition, the line-scan image system is used to detect the dynamic deformation.3-D profile measurement of a moving object using a novel phase-shifting technique is introduced. Digital gratings with two steps phase-shifting are projected periodically onto a measured object surface. The deformed fringe patterns are captured by a frame CCD camera within a short exposure time. By synchronizing the projector and the CCD camera accurately, there is an overlapping part which is the same part of the object among three neighbouring frames. Hence the intensity values at the same surface point modulated by three neighbouring gratings can be obtained, and its phase value can be computed by an improved phase-extracting algorithm. The profile of a specimen is detected by the proposed method. Experimental results demonstrate that this method is effective for the profile measurement of a moving object with larger surface slope coefficient.The phase errors caused by intensity error, such as quantization error, nonlinear response error and saturation error, in projection grating phase-shifting profilometry were discussed through simulation and experiment. (1)The formula of the standard of phase error caused by quantization is deduced, and discrete digital sinusoidal series are used to simulate the phase error. Simulative results indicate that the phase error will be decreased with the increase of the phase-shift steps, the intensity quantization level, the effective quantization coefficient and the modulation depth of the sinusoidal fringe patterns. (2)The phase errors and their periodic wave behavior caused by the nonlinearity of the system are studied. The theoretic formula of the phase error is deduced, from which it is found that the phase error caused by the nonlinearity of the imaging and projected system can be effectively decreased by using the four-step algorithm instead of the three-step algorithm. The experimental results are fit well with the theoretic analysis. (3)The intensity saturation of fringe patterns can also induce the fluctuating phase error. To decrease the phase error introduced by saturation, a novel phase recovering algorithm is proposed and further studied. The real phase can be recovered by the unsaturated intensity values. Simulative results indicate that the phase error caused by the intensity saturation can be effectively decreased by the phase-recovering algorithm when the saturation coefficient of fringe patterns is within the applicability range of the corresponding phase-recovering algorithm. Furthermore, the applicability range of the phase-recovering algorithm will be extended with the increasing of phase-shifting steps.A line-scan image system is constructed to measure the surface profile of a moving object with a variable speed. By using the speed coder, the relative moving speed is obtained and the pulse signal is fed back to the computer for synchronizing the moving object and the line-scan CCD camera. So that, undistorted fringe patterns can be captured by the line-scan camera. The classical fringe skeletonizing method and Fourier transform method are used to evaluate the fringe deformation. Comparing the results obtained by the two different methods, it is found that the measuring accuracy of Fourier transform method is higher than that of fringe skeletonizing method. However, fringe skeletonizing method has the advantage in measuring the surface with higher slope coefficient. Furthermore, the projection dual-frequency composite grating technique is used to solve 2πphase ambiguity problem because of some bigger surface steps. Experimental results are presented to prove the feasibility of the inspection system. Dual-frequency grating inspection can increase the detection accuracy and overcome the 2πphase ambiguity at the same time. The line-scan CCD camera is used to measure vibration characteristics of a string. The CCD line array sensor is set parallel to the vibration direction of a tensile string. There is a distinct track at the center of the captured image because of the different reflectivity between the string and the background. The fundamental frequency of the vibrating string can be obtained from the acquired images by Fourier transform technique. In addition, the line-scan image system is also used to detect the dynamic profile of human arm caused by muscle deformation. One line of the detected object surface is captured along a period of time. Fourier transform method is used to get the profile sequence of the same surface line in different time.更多还原