中文摘要：

设计并实现了一种基于单数码相机的板料成形应变测量系统MOSAS,并对其中的关键技术作了深入研究和讨论。在基于多视角几何原理、编码元技术实现各个拍摄视角相机姿态定位的基础上,本文使用光束平差等全局优化技术进一步提高定位精度;针对方形网格图像,综合运用数字图像处理技术自动实现亚像素精度的图像网格节点提取和网格拓扑关系稳健建立;结合极线约束和网格拓扑,实现多视图下同名网格节点自动鲁棒匹配;提出多视图下的三维重建方法以获得网格节点空间坐标;最后由有限应变理论解算出应变值.通过对网格边长为1mm标准模板的测量实验,表明MOSAS系统的三维重建精度为0.004mm（2σ）,应变准确度不大于0.4%（2σ）。

英文摘要：

A sheet metal forming strain measurement system, namely MOSAS, was designed and implemented, which is based on analyzing the images taken by one single hand-held digital camera. The key techniques concerned are studied and discussed in detail. On the basis of locating the camera position and orientation at every shot realized using multi-view geometry theory and coded target techniques, global optimization algorithms such as beam adjust- ment and etc. are used to realize improved location precision. With integrated use of digital image processing on square grid images, precise extraction of sub-pixel precision level coordinate of grid nodes and topological relationship robust establishing between grid nodes are automatically implemented. Corresponding grid nodes in different views can be automatically and rapidly matched using topological relation and epipolar constraint. Then a multi-view 3D reconstruction method is proposed to obtain space coordinates of grid nodes. Finally, strain can be calculated using finite strain theory. Experiment on a standard template with grid gauge length of 1 mm shows that the 3D reconstruction precision of MOSAS is 0.004 mm （2σ） and its strain precision is not greater than 0.4% （2σ）.