【作者】 薛婷；

【导师】 叶声华；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2007， 博士

【摘要】 根据现代测量技术的发展,针对日益增长的三维视觉检测需求和应用,本论文详细分析了国内外三维视觉检测技术的发展概况及三种通用系统方案,遵循仪器化过程中高精度、通用性、统一性、灵活方便性以及实用性的思路,围绕三维视觉检测中图像特征提取、视觉传感器标定及测量等若干关键技术,从理论和实践上进行了一系列深入的研究,设计完成了相应的实验论证工作,进一步推进了三维视觉检测技术仪器化和实用化的进程。论文完成的主要工作有:一、针对三维视觉检测技术的需求和应用,概括总结了国内外三维视觉检测技术的发展状况,分析了三维尺寸和形貌视觉检测的三种通用系统方案,并对其中的关键技术,尤其目前亟待解决的共性化、关键性技术问题进行了详细阐述;二、综合主动、被动视觉检测法,提出了一种基于立体视觉的结构统一的多功能立体视觉传感器,该传感器模块化、多功能、通用性好,可以实现不同测量特征(孔、棱线、曲面等)的三维视觉检测任务;三、详细分析了各类摄像机标定方法,研究了一种通用的、对摄像机初始参数及标定靶标姿态无严格要求的高精度摄像机标定方法,可以标定出包括摄像机径向畸变和切向畸变在内的内、外部参数;四、针对结构统一多功能立体视觉传感器,建立了统一的标定数学模型,提出了一种基于任意位姿平面靶标的高精度标定方法,通过实验验证,该方法灵活方便、快速简单、标定精度高;五、针对适用于机器人柔性三维视觉检测系统的线结构光视觉传感器,研究了一种基于交比不变性及任意位姿平面标定参照物的高精度现场标定方法,极大地简化了标定设备,灵活方便,精度高;六、研究了三维视觉检测中高精度图像特征提取方法,提出了类椭圆自动分割和中心特征自动提取方法,在详细分析各类光条提取算法的基础上,研究了基于Hessian矩阵的高精度光条(类线条)提取方法,并进行了相关实验验证;七、提出了双目立体视觉传感器可修复的概念,实验证明该方法在对光照敏感的三维视觉检测系统中,具有良好的适应性和精度。八、提出了线结构光视觉传感器测量圆(类圆)孔中心两步法,突破了传统圆(类圆)孔中心测量采用双目传感器的方式,实验验证该方法切实可行,并给出了详细的误差分析。更多还原

【Abstract】 The three-dimensional visual inspection technologies in the world and three universal vision system schemes are exhaustively analyzed in the thesis based on the demands and applications of the technology through the development of the modern measurement technology. The research is mainly conducted to the key technologies of three-dimensional visual inspection, such as image feature extraction, vision sensor calibration and measurement, abiding by the instrumentation regulation of high precision, universality, uniformity, flexibility and practicality. Both theoretical and practical researches and discussions were taken, and some creative theories and means, as well as experimental validation were finished. The main work of the thesis is as follows:1. Introduction of the demands and application domains of three-dimensional visual inspection, the development of three-dimensional visual inspection technologies in the world. Analysis of three universal vision system schemes for dimension and surface visual inspection. Explanation of the common and key technologies, especially not have been solved exhaustively.2. A novel structure-uniform stereovision sensor for various three-dimensional feature inspection tasks, such as hole, edge, surface and so on, based on the active and passive visual inspection methods.3. Analysis of camera calibration methods in detail. The research on a universal camera calibration method with unknown camera initial parameters and planar pattern position, which estimate all the camera parameters including lens radial and tangential distortion coefficient with high precision.4. High-precision calibration solution of a structure-uniform stereovision sensor with free-position planar pattern. The uniform mathematical model was constructed. The solution was proven by experiments to be easy, flexible, with high precision.5. High-precision calibration solution of a structured light stripe vision sensor, which is suitable in flexible three-dimensional visual inspection system based on the industrial robot platform. The solution simplified the calibration set-up greatly, and was proven by experiments to be easy, flexible, with high precision.6. The research on the image feature extraction technology in the visual inspection. The auto-segmentation and sub pixel auto-location method of ellipse was presented. The high-precision light stripe location technology base on the Hessian matrix through exhaustive analysis of light stripe extraction algorithm was proposed and validated by experiment.7. Creation of the reparability method of stereovision sensor. The method was proven by experiment to be practical in the three-dimensional visual inspection system, which is sensitive to the light.8. Creation of the two-step hole inspection technology with structured light stripe vision sensor, and exhaustive analysis of measurement error. The method broke the traditional hole inspection method with stereovision sensor, and was proven by experiment to be valid.更多还原