数字图像相关方法是上世纪八十年代初发展的一种非接触式全场光学测量方法。与目前实验力学中其它几种基于相干光波干涉的重要光测方法（如全息干涉、散斑干涉、云纹干涉等）相比，它具有一些其它光测方法不可比拟的优势。因而该方法适合于工程现场测量、并可直接与各种高空间分辨率的显微观测平台结合实现各种空间尺度下的变形测量。可以说，数字图像相关方法是当前最活跃也最有生命力的光测技术之一，已经在科研和工程上获得无数成功的应用，并展示了广阔的应用前景。目前数字图像相关方法在位移、应变测量精度以及计算速度上还存在一些不足，一些关键问题和关键技术还没有得到充分的认识和解决。本文对二维和三维数字图像相关方法中的一些关键技术进行了深入研究，研究主要集中在如何提高位移和应变测量精度上，例如：（1）如何提高位移测量的精度；（2）如何提高应变测量精度；（3）三维数字图像相关方法中的双目立体视觉系统高精度标定和高精度立体匹配的实现。并在理论研究的基础上，用软件实现了基于数字图像相关方法的高精度变形测量。本文取得的主要研究成果有：（1）对数字图像相关方法中提高位移测量精度的各种亚像素位移测量算法的性能进行了详细研究，给出了明确的结论；（2）提出一种基于人机交换的变形初值估计方法，使数字图像相关方法在被测物体存在较大刚体转动和大变形时仍能进行准确的变形测量（3）推导了数字图像相关方法位移测量精度的理论模型，并基于该模型提出了一种选择图像子区大小的方法；（4）为消除成像镜头畸变对位移测量的影响，提出一种改进的成像镜头畸变估计和校正的方法；（5）针对数字图像相关方法中的应变场测量这一关键技术，提出了基于位移场逐点最小二乘拟合的全场应变测量方法；（6）将最新的双目立体视觉系统标定技术和基于二次形函数的数字图像相关匹配方法结合，实现了高精度的三维数字图像相关方法，并成功地应用于复合材料卫星天线表面的形貌测量和浮空飞行器蒙皮材料在鼓膜实验中的变形测量。
Digital image correlation technique (DIC) is a non-contact full-field optical measuring technique, which was originally developed in the 1980’s of last century. Compared with other optical methods based on the interference of coherent light wave (e.g., Holography Interferometry, Speckle Interferometry and Moire Interferometry), the DIC technique possesses many remarkable advantages. Especially, it is suitable for the deformation measurement at engineering site and can be directly used in combination with various kinds of high-spatial-resolution Microscopic observation platform. It can be said that DIC technique is the most active and most promising one of the existed optical metrologies. It has obtained numerous successful applications in engineering and scientific research, and also demonstrates broad application prospects.However, the DIC technique still exist many deficiencies in terms of displacement measurement accuracy, strain measurement accuracy as well as calculation speed. Some of the key issues and key technologies of DIC have not yet been fully understood and addressed. In this thesis, a number of key issues of 2D and 3D DIC technique have been studied, and efforts were mainly focused on the improvement of displacement and strain measurement accuracy, for instance: (1) how to improve the displacement measurement accuracy by matching the reference and target subsets in DIC technique; (2) how to improve the strain measurement accuracy in DIC technique; (3) how to acquire high-precision calibration of the stereo vision system and high-precision stereo matching in 3D DIC technique.The main achievements of this thesis are listed as following: (1) the performance of various existing sub-pixel registration algorithm has been investigated thoroughly in terms of registration accuracy and computational efficiency; (2) considering the possible cases where large rotations and/or strains may occur between successive images, an technique for initial deformation guess based on human-computer exchange is developed; (3) The theoretical model of displacement measurement of DIC is derived, and an algorithm for optimal selection subset size is further presented based on this model; (4) An improved lens distortion correction technique is presented; (5) A simple yet effective technique based on the least-square fitting of local displacement field is proposed for strain estimation; (6) By combining the up to date stereo vision calibration technique and stereo matching technique using second-order shape function, high-precision 3D DIC technique is realized and is successfully applied to the shape measurement of satellite antenna and deformation measurement of composite film in bulge test.
