由图像序列测量三维物体的方法

The project is aimed at Eucliean reconstruction and the metrology of 3D object from image sequence captured by a hand-held camera. It includes: a) linear and non-linear algorithms to estimate the fundamental matrix with high accuracy and robustness; b) dense matching algorithm which can be applied to image pairs without rectification; c) a novel camera self calibration method based on linear model using no reference objects; d) analysis of non-linear errors caused by the optical system of the camera, which offers a reliable error model for the vision metrological system to achieve a highly accurate 3D reconstruction; e) iterative optimization based on the absolute quadric; f) development of a novel non-contact metrology which makes use of 3D Euclidean reconstruction from 2D image sequence captured by a hand-held camera. A serials of important results have been achieved. 14 academic papers have been published on this subject, among which 6 papers have entered SCI and 7 papers have entered EI. And this project has won the first class Science and Technology Advancement Award from Educational Bureau, Shaanxi Province.

本项目研究由手提相机所获取的三维物体图像序列进行物体的欧氏重建与测量方法。其中包括：手提相机的自校准、图像间亚像素级匹配、图像间几何约束的精确估计、测量误差的统计模型和高精度测量系统方案。本项目所研究的非接触式光电测量方法具有方便、高精度和自动化程度高等优点。研究结果在工业测量、三维通信和机器视觉等方面具有重要应用前景。