【作者】 许丽；

【导师】 张之江；

【作者基本信息】 上海大学 ， 信息与通信工程， 2009， 博士

【摘要】 实体表面的三维数字化技术在数字媒体、虚拟现实、文物保护等领域有广泛的应用背景。论文研究了被测实体在无约束运动(慢速)下的表面密集三维信息获取的原理与技术。在实时编码光设计、提高三维数据获取的准确性(即结构光系统参数的标定、结构光系统误差传递模型的建立与结构优化),以及多场景数据的在线配准等方面开展了深入的研究。在实时编码光设计中,根据条纹偏移机理,提出了编码光设计的空间周期性准则,推导了空间周期性准则应用的约束条件;建立了测量分辨率与测量实时性之间的对应关系。根据空间周期性准则,本文提出了一种新的实时编码光设计方案——周期时空条纹编码。该编码方式既保证了系统的测量分辨率,又实现了对动态场景的三维信息获取。结合结构光系统的光平面模型,提出了基于平面靶标的结构光标定方法,并对其中的计算过程进行了鲁棒性分析。该标定方法可以降低图像处理过程所引入的误差,减少由多次交比不变计算标定点所带来的累积误差,提高了标定点的获取精度,从而实现结构光系统参数的标定。通过对结构光系统的误差分析,建立了结构光系统的误差传递模型;根据该模型推导出三维数据获取精度与标定点样本的提取精度之间的数学关系。并从系统模型对图像提取误差的敏感程度进行分析,优化了结构光系统的结构设计。针对结构光系统所得到稀疏三维数据的特点以及在线配准算法的要求,提出了改进的最近点迭代算法。该算法以刚体变换不变量(四面体体积)为搜索物理量,利用四面体体积的特性,实现了最近点迭代算法在实时性和自动初始估计方面上的改进,从而完成了多场数据的在线配准。系统实验以实际物体表面轮廓的三维信息获取结果为例,验证了本文所研究的动态场景的结构光三维信息获取系统的可靠性,以及本文所提出原理应用到动态场景的三维信息获取中的有效性。更多还原

【Abstract】 Technologies about digital 3D models of real object surface have some applications, such as entertainment, virtual reality, cultural heritage, and so on. In the paper, denser 3D data is acquired when object is slowly moving without limit. Main innovations in the paper include real-time coded pattern design, 3D data acquisition (system calibration, error transmission model and layout optimization), and sparse range image registration.According to mechanism of deformations of stripe, spatial periodicity theory is proposed in the paper, the relation between resolutions and real-time of system is estimated in the paper. Combination with time-space coding, a new real-time coding is designed making use of spatial periodicity, which is called as alternatice time-space stripe coding. It can ensure measuring resolution and accuracy, and acquire the real-time data of dynamic scenes.Calibration method using coplanar target is proposed in the paper. The robustness during calibration is analyzed. The method can obtain some calibration points with high accuracy, and decrease the error from image processing and multiple cross-ratio computation.Error transmission model is proposed in the paper. Based on the model, the relationship between measuring accuracy and accuracy of calibration sample is concluded. In order to obtain higher measuring with layout optimization, some analysis is measuring accuracy robustness to image extraction. Some conclusions are obtained, which can guide system design.Improving iterative closed point algorithm is proposed, which is used for on-line sparse range image registration. In the algorithm, invariant is added during search corresponding points such as the volume of tetrahedron. The volume of tetrahedron is proved as an invariant in the paper. By the characteristics of volume, the algorithm is advanced in the respects of real-time and intial estimation. Then the advanced algorithm completes the on-line data registration from multipleview. Experiments of real object’s surface prove that the relability of real-time structured light system for dynamic scenes, and the validity of some new theories used in real-time structured light system.更多还原