【作者】 关丛荣；

【导师】 于晓洋；

【作者基本信息】 哈尔滨理工大学 ， 测试计量技术及仪器， 2009， 博士

【摘要】 非接触三维测量技术以其效率高、自动化程度高、造价较低等优点,被广泛地应用于工业生产和现实生活中。结构光方法是具有较高实用性和发展潜力的三维非接触测量技术。编码结构光法以其高效、高速和勿需扫描而成为结构光三维测量技术的发展方向。彩色编码较之灰度编码效率更高,相对于空间和直接彩色编码方法,时间彩色编码方法具有较高的准确度和采样密度而成为彩色编码结构光发展方向。提高准确度是测量领域的永恒追求,提高采样密度则是三维视觉测量中极具挑战性的任务。为此,本文以提高时间编码结构光三维测量的准确度和采样密度为目的,研究条纹分隔颜色格雷码结构光三维非接触测量原理及其关键技术。为提高编码结构光的准确度,本文从分析传统格雷码条纹中心解码原理误差出发,提出颜色边缘格雷码解码方法。该方法以彩色条纹边缘作为采样点,从原理上消除了条纹中心解码带来的0.5个最低位固有量化误差;边缘解码与其高位图像中红蓝条纹内部而非边缘相对应,大大降低了格雷码码值被误判的几率、提高解码可靠性;彩色条纹较之灰度条纹还能减少条纹扩散影响产生的边缘检测误差。为提高编码结构光的采样密度,基于颜色边缘格雷码在每相邻彩色条纹处嵌入其它颜色的宽度固定条纹,将宽度固定条纹的边缘和中心作为采样点,达到不增加投射幅数的前提下提高采样密度的目的。采用条纹分隔颜色格雷码编码方法进行三维测量时,作为采样点的彩色条纹边缘和中心的准确检测是保证系统测量精度的关键。针对编码条纹特点,基于行扫描技术,提出基于彩色分量交点亚像素检测条纹边缘算法,以及彩色分量重心法亚像素检测条纹中心算法,并用相关仿真实验初步验证了方法的可行性和准确性。因受系统硬件设备、环境光强度、拍摄角度及空间物体几何形状等因素影响,彩色条纹图像会产生颜色失真影响条纹检测精度,需要进行颜色校正。为此,本文基于自然光环境下对投影仪和摄像机进行颜色标定,用获得的硬件颜色标定矩阵校正所拍摄的彩色条纹图像,以消除图像各彩色分量之间的影响。基于归一化彩色模型l1l2l3空间实现彩色条纹图像归一化,以消除环境光强度、拍摄角度及空间物体几何形状等因素影响,从而提高彩色条纹边缘和中心亚像素的检测精度。采用3dmax软件仿真投影仪、摄像机、被测物和被测环境,所获得的仿真编码图像利用Matlab软件根据给出的系统数学模型重构被测物表面。针对不同仿真三维物面进行了颜色格雷码中心编解码和条纹分隔颜色格雷码编解码方法的对比实验。基于条纹分隔颜色格雷码编码方法组建实验装置,进行系统参数设计,采用线性标定法标定实验装置参数,并对系统不确定度进行了分析,针对典型表面和复杂表面进行三维测量实验。实验结果表明:在系统量程范围内,最大测量误差小于2mm,相对误差小于0.2%;重构的复杂表面与被测表面形状相符,能够较平滑细致地反映被测表面形貌。本文的研究为结构光提供了新的原理和技术,实现了高准确度密集采样三维测量,对提高结构光三维测量技术水平和拓展其应用领域有重要意义。更多还原

【Abstract】 Non-contact 3D measurement technology is widely used in industry and real life for its advantages of high efficiency, high automatization and low cost. Structured light as a kind of non-contact 3D measurement technology has high practicality and developing potential. The coding structured light method which has the advantages of high efficiency, high speed and no scanning is considered as the developing direction of structured light 3D measurement technology. Color coding has much more reliability than grey coding. Compared with the space coding method and direct coding method, time coding method for its advantages of higher accuracy and high sampling density becomes the research direction of color coding structured light. Improving accuracy is the eternal pursuit of the measurement field, and improving sampling density is a highly challenging task in the 3D vision measurement. Therefore, this paper studies time color coding structured light 3D non-contact measurement technology and experimental device to improving accuracy and sampling density.To improve the accuracy of coding structured light, from the error analysis of traditional Gray code center decoding principle, the method of color edge Gray code decoding is proposed, which takes color stripe edge as sampling point and eliminates the 0.5 least significant bit inherent quantization error from stripe center decoding. Edge decoding is corresponding with the inside of red/blue stripe in high-order image, not with the edge of red/blue stripe in high-order image, which reduces misjudge probability of the Gray code value greatly, and enhances decoding reliability. In addition, contrast to grey stripe, color stripe can reduce the edge detection error from stripe diffusion.To improve sampling density of the code structured light, insert different color stripe of specific breadth between adjacent colorful stripes based on color edge Gray code, sample on the edge and center of specific breadth stripe, which can improve sampling density withnot adding project patterns. Adopting the coding method of stripe dividing color Gray code to 3D measurement, the key factor of ensuring measurement precision is detecting the edge and center of color stripe taken as sampling dot accurately. For the character of code stripe and line scan technology, sub-pixel stripe edge detecting algorithm based on color component intersecting point and sub-pixel stripe center detecting algorithm based on color component gravity model method are supposed. And relative simulation experiments have been made to validate the feasibility and veracity of the method.Considering the factors of system hardware, ambient light illumination, shooting angle and space object geometry shape, color distortion of color stripe image from these factors can impact stripe detection precision. So it is necessary to revise color. In this paper, color calibration to projection apparatus and camera under natural light is adopted to remove the influences from different color components. Normalized color model l1l2l3 space realizes the normalization of color stripe image, which can remove the influences from the factors of ambient light illumination, shooting angle and space object geometry shape, and ensure detecting the edge and center of color stripe on sub-pixel order accurately. Simulate projection apparatus, camera, measurement object and environment, and the simulation code images have been used to reconstruct object surface under specific math model with 3dmax and Matlab. Contrast experiments between color Gray code center coding, decoding and stripe dividing color Gray code coding, decoding.It constructs the experiment devices based on stripe dividing color Gray code coding method, and designs the system parameters, calibrates parameters of experiment devices using linear calibration. Uncertainty of the measurement has been analyzed and 3D measurement experiments for typical and complex surfaces have been made. From the results of the experiments, in the measurement range, the maximum measurement error is less than 2mm, measurement accuracy is 0.2%, and reconstructed complex surface can match the measurement surface and reflect the surface pattern smoothly and particularly.This research provides new principle and technology for structured light, realizes 3D measurement under high accuracy and dense sampling, which is important to improve the structured light 3D measurement technology level and enlarge application domain.更多还原